eprintid: 18896
rev_number: 15
eprint_status: archive
userid: 1888
dir: disk0/00/01/88/96
datestamp: 2015-07-01 11:10:58
lastmod: 2015-07-09 08:43:25
status_changed: 2015-07-01 11:10:58
type: doctoralThesis
metadata_visibility: show
creators_name: Kreplin, Katharina
title: A Novel Flavour Tagging Algorithm using Machine Learning Techniques and a Precision Measurement of the B0-AntiB0 Oscillation Frequency at the LHCb Experiment
subjects: ddc-004
subjects: ddc-500
subjects: ddc-530
divisions: i-130200
adv_faculty: af-13
keywords: Flavour tagging;machine learning;artificial neural network;precision measurement;LHCb;LHC;Bd mixing;oscillation frequency
cterms_swd: machine learning
cterms_swd: artificial neural network
cterms_swd: LHCb precision measurement
cterms_swd: flavour tagging
cterms_swd: Bd mixing
cterms_swd: oscillation frequency
cterms_swd: LHC
abstract: This thesis presents a novel flavour tagging algorithm using machine learning techniques and a precision measurement of the B0-AntiB0 oscillation frequency delta m_d using semileptonic B0 decays. The LHC Run I data set is used which corresponds to 3 fb^-1 of data taken by the LHCb experiment at a center-of-mass energy of 7 TeV and 8 TeV. The performance of flavour tagging algorithms, exploiting the b Anti-b pair production and the b quark hadronization, is relatively low at the LHC due to the large amount of soft QCD background in inelastic proton-proton collisions. The standard approach is a cut-based selection of particles, whose charges are correlated to the production flavour of the B meson. The novel tagging algorithm classifies the particles using an artificial neural network (ANN). It assigns higher weights to particles, which are likely to be correlated to the b flavour. A second ANN combines the particles with the highest weights to derive the tagging decision. An increase of the opposite side kaon tagging performance of 50% and 30% is achieved on B^+ to J/Psi K^+ data. The second number corresponds to a readjustment of the algorithm to the B0_s production topology. This algorithm is employed in the precision measurement of delta m_d. A data set of 3.2x10^6 semileptonic B0 decays is analysed, where the B0 decays into a D^-(K^+ pi^- pi^-) or D^*- (pi^- AntiD0(K^+ pi^-)) and a mu^+ nu_mu pair.
The nu_mu is not reconstructed, therefore, the B0 momentum needs to be statistically corrected for the missing momentum of the neutrino to compute the correct B0 decay time. A result of delta m_d = 0.503 +- 0.002 (stat.) +- 0.001 (syst.) ps^-1 is obtained. This is the world's best measurement of this quantity.

date: 2015
id_scheme: DOI
id_number: 10.11588/heidok.00018896
ppn_swb: 1657366979
own_urn: urn:nbn:de:bsz:16-heidok-188967
date_accepted: 2015-06-10
advisor: HASH(0x561a629d2110)
language: eng
bibsort: KREPLINKATANOVELFLAV2015
full_text_status: public
citation:   Kreplin, Katharina  (2015) A Novel Flavour Tagging Algorithm using Machine Learning Techniques and a Precision Measurement of the B0-AntiB0 Oscillation Frequency at the LHCb Experiment.  [Dissertation]     
document_url: https://archiv.ub.uni-heidelberg.de/volltextserver/18896/1/2015-03-30_Dissertation-Katharina_Kreplin.pdf