School of Physics - Theses
Permanent URI for this collection
Search Results
1 - 7 of 7
-
ItemAddressing domain shift in deeply-learned jet tagging at the LHC( 2023-09)Over the last fifteen years, deep learning has emerged as an extremely powerful tool for exploiting large datasets. At the Large Hadron Collider, which has been in operation over the same time span, an important use case is to identify the initiating particles of hadronic jets. Due to the complexity of the radiation patterns within jets, neural network-based classifiers are able to out-perform traditional techniques for jet tagging. While these approaches are powerful, neural networks must be applied carefully to avoid performance losses in the presence of domain shift—where the data on which a model is evaluated follows different statistics to the training dataset. This thesis presents studies of possible strategies to mitigate domain shift in the application of deep learning to jet tagging. Firstly, we develop a deep generative model that can separately learn the distribution of quark and gluon jets from mixed samples. Building on the jet topics framework, this model provides the ability to sample quark and gluon jets in high dimension without taking input from Monte Carlo simulations. We demonstrate the advantage of the model over a conventional approach in terms of estimating the performance of a quark/gluon classifier on experimental data. One can also use likelihoods under the model to perform classification that is robust to outliers. We go on to evaluate fully- and weakly-supervised classifiers using real datasets collected at the CMS experiment. Two measurements of the quark/gluon mixture proportions of the datasets are made under different assumptions. Compared to the predictions based on simulation, we either over- or under-estimate the quark fractions of each sample depending on which assumption is made. When estimating the discrimination power of the classifiers in real data we find that while the absolute performance depends on the choice of fractions, the rankings among the models are stable. In particular, weakly-supervised models trained on real jets outperform both simulation-trained models. Our generative networks yield competitive classification and provide a better model for the quark and gluon jet topic distributions in data than the simulation. Finally, we investigate the performance of a number of methods for training mass-generalised jet taggers, with a focus on algorithms that leverage meta-learning. We study the discrimination of jets from boosted Z' bosons against a QCD background and evaluate the networks' performance at masses distant from those used in training. We find that a simple data augmentation strategy that standardises the angular scale of jets with different masses is sufficient to produce strong generalisation. The meta-learning algorithms provide only a small improvement in generalisation when combined with this augmentation.
-
ItemSearching for signals of Dark Matter produced with top quark pairs using the ATLAS detector( 2019)Understanding the nature of Dark Matter is a key goal in modern physics. The observed gravita-tional interactions of galaxies and galactic clusters, along with theories of structure formation in the early universe, indicate the existence of Dark Matter. Evidence of the specific nature of Dark Matter remains elusive however. Particle collider experiments search for evidence of Dark Matter production within energetic proton collisions. One strategy employed in this field is to make minimal assumptions about new particles and couplings to Standard Model particles, in order to explore the range of possibilities without being overly constrained by narrow assumptions. This thesis focuses on the assumption that Dark Matter couples strongly to the heavier quarks, which motivates searching for processes where it is produced in association with pairs of top quarks. An analysis is presented on the 2015 and 2016 “Run 2" dataset taken with the ATLAS detector, consisting of 36.1 fb -1 of proton-proton collisions at the Large Hadron Collider. This analysis studies the hypothesis of Dark Matter production in conjunction with hadronically decaying top quarks. No excess above the estimated Standard Model backgrounds is observed, and constraints on the allowed cross-sections are presented. When making minimal assumptions about the nature of Dark Matter, scalar mediator masses below 20 GeV are excluded. These results are then translated to more specific and complete Two Higgs Doublet models that feature for example in Supersymmetry that also predict the same final states, and constraints on the parameter space of these models are presented.
-
ItemSearch for the Higgs boson in the WH production mode with H→WW* decay using the ATLAS detector( 2018)In order for fundamental particles to attain mass, the electroweak symmetry of the Standard Model (SM) of particle physics must be broken. The simplest way of breaking this symmetry is via the SM Higgs Mechanism, and it predicts the existence of a new particle called the SM Higgs boson. This particle should be experimentally accessible at the high-energy frontier, and so its discovery is considered as one of the most important goals in modern particle physics. This goal was partially achieved in July 2012, when the ATLAS and CMS experiments at the Large Hadron Collider (LHC) announced the discovery of a Higgs-like particle with a mass of around 125 GeV. The next step for both experiments is to scrutinize the properties of this new particle in order to determine whether it is, in fact, the SM Higgs boson. One high-priority objective of the ATLAS experiment is to confirm the SM predictions for Higgs boson production at the LHC. The main result presented in this thesis contributes to this objective by searching for the WH production mode with subsequent H→WW* decay. The H→WW* decay channel provides a sensitive probe of WH production due to its large branching ratio and clean detector signature. Moreover, this signal process provides important information on the Higgs boson couplings, since it only involves couplings to W bosons at both the production and decay vertices at tree-level. The search for this signal process was conducted using LHC proton-proton collision data collected by the ATLAS detector. This data was recorded at centre-of-mass energies of 7, 8 and 13 TeV corresponding to integrated luminosities of 4.5, 20.3 and 5.8 fb−1, respectively. The analysis strategy was first developed using the 7 and 8 TeV data samples and it was used to measure the relative signal strength with respect to the SM expectation. For a Higgs boson of mass 125 GeV, the observed value of the signal strength was determined to be 0.72 +1.2−1.1 (stat.) +0.4−0.3 (sys.). The analysis method was then extended to perform the first measurement of the signal strength at 13 TeV, with this quantity measured to be 3.2 +3.7−3.2 (stat) +2.3−2.7 (sys). An overall excess was observed at 0.66σ (0.77σ) significance with respect to the background-only hypothesis in the 7+8 TeV (13 TeV) data. All measurements are in agreement with the SM expectations.
-
ItemHiggs boson and new physics searches in events with same-sign lepton pairs at the LHC( 2017)The physics of elementary particles is well described by the Standard Model of particle physics. This theory predicts with great accuracy the electroweak and strong interactions of elementary particles. The introduction of the Brout-Englert-Higgs mechanism in Standard Model has also given justification to the non-null mass of the elementary particles. This mechanism has been validated by the discovery of the Higgs boson and the measurement of its properties. Despite the enormous and broad success of the Standard Model, a few observations cannot be explained by this theory. For instance, neutrino oscillations would require the neutrinos to have non-null mass in contrast with the Standard Model prediction. Several solutions to this problem have been proposed, but none has been validated yet. Some of these theories have additional modes to produce lepton pairs with same electric charge (same-sign lepton pairs). This production is also expected to be enhanced in other theories beyond the Standard Model. The focus of this thesis is the inclusive search for anomalous production of same-sign lepton pairs in proton-proton collisions. The specific case of resonant production, typical of the Type-II see-saw and Zee-Babu models of neutrino mass generation, is also considered. 20.3/fb of data collected by the ATLAS experiment at a centre-of-mass energy of 8 TeV is analysed. No significant production above the Standard Model prediction is found, thus the results are used to set upper limits on the cross-section of processes beyond the Standard Model. Limits are placed as a function of the di-lepton invariant mass within a fiducial region defined according to the event selection criteria. Exclusion limits are also derived for specific models of resonant production. The production of same-sign lepton pairs from proton-proton collisions occurs rarely in the Standard Model. One process which generates same-sign lepton pairs is the associated production of a Higgs boson and a top anti-top quark pair (ttH). The cross-section of this production mode is proportional to the Top Yukawa coupling squared, while other processes are only sensitive to the combination of this constant with several other couplings. The measurement of the ttH cross-section therefore constitutes an important test of the Standard Model and the Brout-Englert-Higgs mechanism. The ttH production has never been observed before, and is studied in this thesis. The analysis uses collisions at a centre-of-mass energy of 8 TeV collected by the ATLAS experiment containing a same-sign lepton pair and jets. The measured production cross-section is 2.8+2.1−1.9 times the value expected in the Standard Model. This result has been combined with other analyses to obtain a more precise measurement of the ttH cross-section and the Top Yukawa coupling.
-
ItemMono-X searches for simplified models of dark matter( 2016)The identity of dark matter remains one of the big open questions in particle physics; while much is known about its distribution throughout the Universe, very little is understood about its particle nature. In particular, a small but non-zero coupling to the Standard Model (SM) sector has not yet been ruled out. WIMP-type dark matter (DM), with weak-scale mass and couplings, may therefore be produced in proton collisions with the Large Hadron Collider (LHC), and detected by the ATLAS experiment. Several collider searches are presented, which utilise the mono-X+ MET (missing transverse energy) topology, wherein DM (the presence of which is inferred through the observation of missing transverse energy) is produced in association with some object X. The mono-jet process has the largest cross section, however mono-boson analyses, the focus of this thesis, have other advantages. The mono-Z(l+l−) channel benefits from the straightforward identification of charged leptons within the detector and removal of the multi-jet background, while the mono-W/Z(jj) channel is able to utilise the growing collection of electroweak boson identification techniques which exploit the two-prong substructure of a large-radius jet. This thesis describes two ATLAS analyses that seek to constrain both Effective Field Theory (EFT) models and simplified models of DM. The ATLAS mono-Z(ll) analysis uses 20.3 fb−1 of data produced at 8 TeV and selects events with a leptonically-decaying Z boson produced back-to-back with a large amount of MET. A cut-and-count method finds that no excess above the SM prediction is observed, and so constraints are calculated for the suppression scale Λ of the EFTs, and for the quark-DM-mediator coupling of a simplified model with a scalar mediator exchanged in the t-channel. The ATLAS mono-W/Z(jj) analysis uses the first 3.2 fb−1 of data produced at 13 TeV, and selects events with a single large-radius jet produced in association with MET. A profile likelihood fit of the SM background estimation and data is used to extract a limit on the signal strength for a vector mediator s-channel simplified model, and converted to a limit on the suppression scale Λ for a ZZχχ contact operator. A reinterpretation of Run I results from ATLAS for three common simplified models is also presented, including a comparison of the results from the mono-jet, mono-Z(l+l−) and mono-W/Z(jj) channels. Limits on the model coupling strengths are discussed. The strongest constraints are obtained with the mono-jet channel, however the leptonic mono-Z channel is able to remove the large multi-jet background to attain limits that are weaker by only a factor of a few. It is essential that the reconstruction of objects within the ATLAS detector, along with their energy measurement and calibration, is well understood and that the performance is optimised. Along with a general discussion of the relevant objects in the detector (leptons, jets and MET), the in situ measurement of corrections to the energy scale of hadronically-decaying tau leptons is described.
-
ItemSearching for the Light Supersymmetric Top Quark with the ATLAS experiment( 2011)The nature of dark matter and the source of the matter-antimatter asymmetry in the universe are two of the most important questions in particle physics and cosmology. The current Standard Model of particle physics, while being a very successful description of the observed fundamental particles and their interactions, cannot fully account for either of these phenomena. Theoretical extensions of the Standard Model, however, possibly can. One such extension is the Minimal Supersymmetric Standard Model (MSSM). This thesis begins by exploring the MSSM parameter space in which the matter- antimatter asymmetry of the universe is dynamically generated through electroweak baryogenesis. In this scenario, one of the supersymmetric partners to the top quark, the light stop quark, must be lighter than the top quark. It is found that this parameter space region is highly constrained by experimental limits on the electric dipole moment of the electron and the branching ratio of a bottom quark into a strange quark and a photon. If the additional requirement of matching the observed dark matter abundance by the relic density of the lightest supersymmetry particle is necessitated, the allowed MSSM parameter space is further constrained. The focus of the thesis then moves to the investigation of the collider phenomenology of supersymmetric electroweak baryogenesis, in particular, the evaluation of the discovery potential of light stop quark pair production at the LHC using the ATLAS experiment. This study assumes a light stop decay topology involving the lightest chargino and neutralino where the visible final state products mimic those from top quark pair production. Feasibility studies are performed for proton-proton collisions at centre of mass energies of 10 TeV and an integrated luminosity of 1/fb, concentrating on the dileptonic and semileptonic decay channels where there are two or one charged leptons in the final state. It is found that signal points with stop masses less then 120 GeV and stop-neutralino mass differences greater than 60 GeV have the greatest discovery potential in the dileptonic decay channel, while the semileptonic decay channel is swamped by backgrounds and requires detailed understanding of the detector and backgrounds in order to extract a signal. Finally, a preliminary study is conducted on 41.4/pb of data collected at collisions with centre of mass energies of 7 TeV in the dielectron decay channel, focusing on the understanding of selection variables and backgrounds.
-
ItemDiscovery potential of neutral MSSM Higgs bosons decaying to tau-lepton pairs in the ATLAS experiment( 2011)The Large Hadron Collider (LHC) is the highest energy particle collider ever built. It recently began operation at CERN and will probe physics at unprecedented scales. ATLAS is a particle detector located at one of the collision points on the LHC ring and is designed to be sensitive to the wide range of physics that could be produced. The primary objective of the LHC experiments is to determine the mechanism of electroweak symmetry breaking, of which many theoretical models exist. In the Minimal Supersymmetric Standard Model (MSSM), electroweak symmetry breaking is achieved through the Higgs mechanism, however, the Higgs sector must be extended with respect to the Standard Model and contains five physical Higgs bosons. The discovery potential of the MSSM Higgs bosons in ATLAS has been evaluated in previous studies, demonstrating adequate sensitivity for discovery or exclusion over a large region of the parameter space. However, these studies were performed using now outdated software, without an estimation of the expected systematic uncertainties or the inclusion of data-driven background estimation procedures. In this thesis, the discovery potential of the neutral MSSM Higgs bosons when decaying to tau-lepton pairs in the ATLAS experiment is evaluated. One tau is required to decay leptonically while the other is required to decay hadronically. Higgs boson mass hypotheses in the range 150 GeV - 800 GeV are considered. The study assumes a proton-proton collision energy of 14 TeV and an integrated luminosity of 30/fb. The expected systematic uncertainty on the background measurement is evaluated and included in the calculation of the discovery potential. Data-driven estimation techniques are developed for the W+jets and QCD di-jet backgrounds. The contributions of all signal and background processes are estimated using Monte Carlo simulated event samples. The discovery potential is interpreted in the mh-max benchmark scenario, and is presented in the mA-tan(beta) plane. A small degradation in performance with respect to the previous studies is found for Higgs boson masses below 450 GeV due to the inclusion of systematic uncertainties. It is confirmed that a large fraction of the parameter space will be accessible to the ATLAS experiment, which will be able to probe far beyond the regions already excluded by the LEP and Tevatron experiments. Two separate studies are also included, describing contributions to the modelling of hadronic tau reconstruction in the ATLAS fast simulation packages ATLFAST-I and ATLFAST-II. Firstly, a complete parameterisation of the calorimeter-based hadronic tau reconstruction for use in ATLFAST-I is presented. Secondly, the validation of the track-based hadronic tau reconstruction in ATLFAST-II is presented, including the extraction of correction terms to match the performance in ATLFAST-II to the standard ATLAS simulation.