The neutrino masses and flavor mixing are a direct evidence of new physics Beyond the Standard Model (BSM): the study of neutrino oscillations is thus a royal road to the search of new, unexpected phenomena. In particular, the analysis of neutrino and antineutrino oscillations at T2K and NOVA are providing first exciting hints of CP violation in the leptonic sector. This would be a major discovery related with one of the most fundamental questions in High Energy Physics: why there is an asymmetry between matter and antimatter in the Universe?
T2K is a neutrino experiment designed to investigate how neutrinos change from one flavour to another as they travel (neutrino oscillations). An intense beam of muon neutrinos is generated at the J-PARC nuclear physics site on the East coast of Japan and directed across the country to the Super-Kamiokande neutrino detector in the mountains of western Japan. The beam is measured once before it leaves the J-PARC site, using the near detector ND280, and again at Super-Kamiokande: the change in the measured intensity and composition of the beam is used to provide information on the properties of neutrinos.
The work of the proposed thesis will concentrate on the installation, commissioning and scientific exploitation of the High-Angle Time Projection Chamber (High-Angle TPC). The goal of this new detector is to improve the Near Detector performance, to measure the neutrino interaction rate and to constrain the neutrino interaction cross-sections so that the uncertainty in the number of predicted events at Super-Kamiokande is reduced to about 4% (from about 8% as of today). This will allow improving the physics reach of the T2K-II project. This goal is achieved by modifying the upstream part of the detector, adding a new highly granular scintillator detector (Super-FGD), two new TPCs and six Time Of Flight planes.
The new TPCs will be read out by resistive Micromegas detectors and instrumented with a compact and light field cage. The TPC will measure charge; momentum and directions of tracks produced by charged particles and will provide particle identification through dE/dx measurement with excellent efficiency and precision. Detector prototypes of the new TPCs have been successfully tested in Summer 2018, 2019 and 2021 at CERN and DESY test beams validating the detector technologies and their performance.
The IRFU group is heavily involved in the TPC project, especially in Micromegas detectors production and tests. The detector construction is on going for an installation in Japan in 2022.
The first part of the thesis will be devoted to TPC data analyses. The student will contribute to the commissioning and first beam data taking and analyses foreseen in 2023. The work will focus on the characterization of the resistive Micromegas detector. This is an innovative detector, which will exploit for the first time the resistive technology to improve the resolution on track reconstruction in the TPC. The IRFU group has been initiator of both the original Micromegas technology and of its resistive implementation.
A cutting-edge R&D conducted at IRFU has brought today to the deployment of such technology in a real detector. A seminal, unprecedented work of quantitative understanding and simulation of the charge spread in the resistive detector is on going.
New and sophisticated reconstruction algorithm must be developed to fully profit of the new detector capabilities. In particular, the timing information related with the resistive phenomena and encoded in the signal waveforms should be exploited. Indeed the resistive technology brings improved performances but also new challenges: the charge spread over multiple pads, induced by the resistive phenomena, will highly increase the multiplicity of signals to be analyzed.
Machine Learning (ML) methods will be explored to perform background-rejection and particle ID purposes at the pre-selection stage as well as evaluate them for the pattern recognition stage of track reconstruction. ML are known to have improved performance of many experiments in neutrino Physics (SNO, NEXT, NOvA, KamLAND-Zen, EXO-200, MINERvA). Producing images like structures from detectors data allows to benefit of the pattern recognition capabilities of these tools and enhancing useful features of the data, they can improve not only events but also particles classification tasks.
We propose as a first step to apply ML techniques to treat TPC information. The arrival time on the resistive anode plane gives the z coordinate perpendicular to that (x,y) plane. An event in the TPC is represented by two images projecting on (x,y) and (y,z) planes with the color scale being the pad charge to incorporate dE/dx information. This will allow treating the TPC information as images and to use the powerful ML algorithms used in image analysis. We plan to use implementations relying on Convolutional Neural Network (CNN) (for some, adapting the GoogLeNet CNN architecture) originally designed to image recognition. To significantly reduce training time, Graphical Processing Units (GPUs) will be used, which enable to perform computing operations in parallel. At the TPC level, we aim to use such techniques for Particle identification (PID) and possibly for pattern recognition.
Next we plan to use ML techniques combining TPC and the central SFGD for particle identification (muon from pion and from proton) as well as for event classification task. In the ND280, the beam of muon neutrinos interacts predominantly via the Charged Current Quasi Elastic interaction. For the purposes of the oscillation analysis, data are separated by event topology into one of three categories based on number of final state pions (no pions, one charged pion or any number of pions). A repository will be prepared, which will contain the images in a format suitable for the training of different ML algorithms. The samples defined above can be selected by using available data, collected by T2K. Other charged current event will fall in background sample.
A framework will be developed to allow the testing of various algorithms for object detection and classification.
The second part of the thesis will be dedicated to the analysis of the first T2K neutrino beam data, collected with the ND280 Upgrade detector, in order to extract a new, most precise, measurement of neutrino oscillations. Thanks to the increased statistics and the improved control of systematic uncertainties with ND280 upgrade, the project has the potential to achieve the best worldwide constrains on CP violation in the leptonic sector. The work will focus on the definition of the selection of the new ND280 samples, the evaluation of the corresponding experimental systematic uncertainties and the modification of the analysis framework for the fit to the neutrino oscillation parameters. The extraction of near detector constraints must be deeply modified to include the information of outgoing detected protons and neutrons coming from neutrino interactions and nuclei, which are completely missing in the present analysis. In parallel, the theoretical systematic uncertainties will need to be reevaluated on the basis of the new exclusive models of neutrino-nucleus interactions.