Michael De Nuccio ; Francesco Forti

2018
University of Pisa Pisa

Abstract: The goal of this thesis is the search for an Axion-Like Particle (ALP) A in the decay B± → K±A, A → γγ in the data collected by the BABAR experiment from 1999 to 2007, and the projections to the Belle II experiment, that will soon start collecting data (April 2018). The ALPs are neutral, massive, pseudoscalar particles constituting an extension of the axionic model. The axion was originally hypothesized by Peccei and Quinn to solve the problem of the non-violation of CP, i.e. the symmetry under charge conjugation and spatial parity, by the strong interactions. The strong interaction Lagrangian includes CP-violating terms controlled by parameters that can be interpreted as angles. Without constraints they can assume any value in [0,2π), while the fact that no CP violation has ever been observed in strong processes implies that these parameters must be very close to zero, with need of a fine-tuning of their values. To solve this problem, Peccei and Quinn hypothesized the existence of the axion, a particle that, thanks to a spontaneous symmetry breaking mechanism, would have naturally made these parameters null. Although all the experimental searches carried out to find the axion have been unsuccessful, the axionic model has lead to the creation of a whole family of successors, predicting the existence of Axion-Like Particles, possible candidates to solve other open problems in Physics, as the dark matter. The ALP searched in this analysis would be produced in flavor changing neutral currents processes, rare in the Standard Model, and presents a characteristic signature, with good prospects for discovery. Since there are no constraints from the theory about the ALP mass, the search is carried out up to the kinematical limit, given by M(A)max = M(B±) − M(K±) = 4786 MeV/c^2 . A B-factory is an accelerator optimized to generate a high number of B mesons: the PEP II accelerator, SLAC, California, hosted the BABAR experiment, while SuperKEKEB, KEK, Japan, hosts the Belle II experiment. The BABAR experiment has collected 424.2 fb^−1 of integrated luminosity at the energy of the Υ (4S) meson, and this is the data sample available for my analysis. The objective of this analysis is to reconstruct the decay of a charged B meson into a charged K and two photons, and to search resonances in the diphoton invariant mass spectrum. Various multivariate analysis have been tested to discriminate the signal from the background, and different selection strategies have been taken into account to reject the two main backgrounds, i.e. the one related to B decays and the one related to light quarks production (uds) and quark c. In order to maximize the selection efficiency the rest of the event is not reconstructed. The uds background is jet-like, while the signal has the spherical topology of a B decay, hence the shape variables, that describe the topology of the event, are particularly useful to reject this source of background. The rejection of the B-like background is more complex because of the similarities with the signal. The discrimination is performed with two Boosted Decision Trees, one optimized to reject the B-like background, the other to reject the uds background. Twelve discriminating variables have been used, including shape variables, kinematic variables like mES and ∆E, and a variable related to the particle identification of the K candidate. The multivariate selection is preceded by a loose pre-selection, finalized to reduce the quantity of data to process without losing signal events, to train the selection on a reduced and simplified sample. The selection has been trained on a signal sample produced with Monte Carlo simulation, and modified to have a flat diphoton mass distribution in the considered range: this has been done to avoid the introduction of biases in the training of the selection procedure, caused by the overspecialization over a specific signal ALP mass. In this way it is also possible to include in the selection variables that are dependent on the ALP mass, like the energy of the two photons. The signal yield is finally extracted from a fit to the diphoton invariant mass distribution. The efficiency of the entire selection (pre-selection and BDTs) is between 17% and 19% for the signal (depending on the ALP mass taken into consideration), while the efficiency for the background events of type B and uds is, respectively, 5.0 × 10^−6 and 3.9 × 10^−6. The BABAR experiment collected data during six main Runs: Run 3 has been used as a control sample to verify the validity of the background Monte Carlo samples. Only the Run 3 has been observed as this is a blinded analysis, and the analysts do not have access to the entire data sample before having completed the selection procedure, to avoid the introduction of any bias due to the optimization on the analyzed sample. The diphoton invariant mass distributions for the Run 3 data and for the simulated data are in good agreement, nonetheless the simulation underestimates the total number of events with respect to the Run 3. To avoid biases deriving from an improper modelization of the Monte Carlo, the probability distribution function that models the background has been extracted directly from the control sample. I performed a scan on the control sample, locally fitting the data with a function obtained as a sum of the background and the signal functions, for different ALP mass hypotheses. Assuming that the signal strength is negligible, we expect to find a signal yield distribution that is approximately Gaussian and centered in zero. The next analysis steps will consist in the generation of Toy Monte Carlo to evaluate the performances of the fit procedure and in the quantitative evaluation of the systematic uncertainties. My thesis activity did not only revolve around the analysis for the ALP search, but given the possibility to extend this search also to the upcoming Belle II data sample, it included a collaboration with this experiment. During the course of my thesis I participated to the quality tests performed on the double-sided silicon strip detector modules of the Belle II Silicon Vertex Detector, which were assembled in Pisa. I also contributed to the Silicon Vertex Detector offline reconstruction software, developing and implementing the algorithm to extract the hit time of the particles crossing the detector. Using this time information it is possible to reduce the off-time background and thus improve the tracking performances. The Belle II experiment is foreseen to collect an integrated luminosity equal to 100 times the one collected by BABAR; I also roughly evaluated the sensitivity on the branching ratio of the analyzed channel using the BABAR and the Belle II data samples.

Note: Presented on 19 04 2018
Note: MSc