Neutrinos

Thorsten Lux

For more than a decade, IFAE has been contributing to several key experiments in the field of neutrino physics, such as K2K, which obtained the first measurement of neutrino oscillations with a neutrino beam from an accelerator, and T2K, that presented in 2011 the first indication of the transformation of muon neutrinos into electron neutrinos, thereby demonstrating a non-zero value for the third mixing angle and more recently published the first measurement of the delta CP phase. Currently, the focus of the group lays on the preparation of the T2K near detector upgrade.

T2K Scientific Results in 2020

The T2K Collaboration has published in Nature new results showing the strongest constraint yet on the parameter that governs the breaking of the symmetry between matter and antimatter in neutrino oscillations. Using beams of muon neutrinos and muon antineutrinos, T2K has studied how these particles and antiparticles transition into electron neutrinos and electron antineutrinos, respectively.

The parameter governing the matter/antimatter symmetry breaking in neutrino oscillation, called δcp phase, can take a value from -180º to 180º. For the first time, T2K has disfavored almost half of the possible values at the 99.7% (3σ) confidence level and is starting to reveal a basic property of neutrinos that has not been measured until now. This is an important step on the way to knowing whether or not neutrinos and antineutrinos behave differently.

Figure 2: The published results indicating that the delta CP phase is close to maximal although the possibility of CP conservation is not completely excluded on 3 sigma level.

Exhaustive Neural Importance Sampling Applied To Monte Carlo Event Generation

In modern science and engineering disciplines, the generation of random samples from a given probability density function, p(x), to obtain datasets or compute expectation values has become an essential tool. For the common case that no analytical solution can be found to sample directly from p(x), Monte Carlo (MC) methods are used. A method which can provide good precision is rejection sampling. However, especially for higher-dimensional problems, this method becomes highly inefficient if the proposal function, q(x), the function from which the samples are generated is not a good approximation of p(x). Sebastian Pina-Otey, the first industrial PhD student of IFAE and Grupo AIA, developed a method based on machine learning techniques to find a proposal function which opens the possibility of highly efficient rejection sampling also for high dimensional problems. The method and performance tests were published in Physical Review D 102, 0130.

Figure 3: Performance test of the exhaustive neural importance sampling method for a neutrino cross section: The result for the true distribution, p(x) (Right) The result using rejection sampling with the constructed proposal function q(x)

Figure 4: Performance test of the exhaustive neural importance sampling method for a neutrino cross section: The result using rejection sampling with the constructed proposal function q(x)