Bachelorarbeiten

Investigating the polarization of vector bosons (VB) in VB production processes provides tests for the Standard Model's gauge symmetry structure and the concrete mechanism of electroweak symmetry breaking, as well as sensitivity to new physics. As one of the VB production processes with a very clean signature due to four charged leptons in the final state, the ZZ electroweak production process is very interesting for this purpose. Some literature studies on this process already exist. Since recently, the Monte-Carlo-Eventgenerator Sherpa is able to simulate polarized cross section of vector boson production. The goal of this thesis is to test this new feature with this process. Furthermore, aspects of this process not yet studied, such as the contribution of particle showers, can be investigated.

Investigating the polarization of vector bosons (VB) in VB production processes provides tests for the Standard Model's gauge symmetry structure and the concrete mechanism of electroweak symmetry breaking, as well as sensitivity to new physics. The ATLAS VB polarisation analysis is currently trying to measure the contribution of longitudinally-polarised W-bosons in the scattering of two same-charged W-bosons. To enhance the measurement’s sensitivity, artificial neural networks (ANNs) are used to separate the different VB polarisation components. At the moment, the Monte-Carlo simulations used to train the ANNs are based on the matrix element generator MadGraph.

Since recently, the event generator Sherpa is also able to simulate polarized cross sections of VB production processes. In preparation for its application in the ATLAS VB polarisation analysis, a first proof of concept study has been performed to investigate different methods to train ANNs to separate the VB polarisations. However, these investigations are so far based on “truth-level” simulations, i.e. without the inclusion of detector effects. In addition to extending the existing study to more complex training procedures, the goal of this Bachelor thesis is to apply the collected knowledge to “reconstruction-level” samples that include the effects of the detector simulation and object reconstruction. This allows a direct comparison of the expected measurement sensitivity when using ANNs trained on MadGraph vs. Sherpa-samples.

Triboson production V(->ll)V(->ll)V(->qq) is a component of the signal definition in vector boson scattering, but is sometimes ignored or generated as a separate process. This thesis should study the impact of the triboson component. In particular the effect of dim-8 EFT operators in the quartic vertex should be studied to quantify the effect of ignoring the triboson production in an EFT study.

In the Standard Model Effective Field Theory expansion operators of dim-6 can affect triple gauge couplings. Typically in VBS analysis dim-6 operators are not considered for EFT interpretations as they are better constrained on inclusive diboson analysis. Nevertheless, on the electroweak production of WZjj, diagrams with triple coupling play a role. In this thesis, we would like to quantify the possible effect of the dimension-6 operators when they are evaluated at their best limit value in the WZjj phase space. We will study as well the sensitivity of the WZjj to set limits on dim-6 operators by performing a fit to the data in this region only.

For the WZ fully leptonic resonant search the invariant mass distribution is used to look for a bump. Nevertheless in order to build the invariant mass using detector level information some assumptions are needed. One particularly important assumption we made is the W bosón is on-shell to be able to resolve the issue of the missing neutrino information. In this thesis, we would like to have a study of the impact of this assumption on the resonant mass reconstruction by comparing the detector reconstructed mass and the truth simulated mass.

The simulation of different polarisation states allows us to study effects from physics beyond the Standard Model in more detail. In this project we want to study, how effective-field-theory extensions of the Standard Model will affect different polarisations of W and Z bosons in WZ pair production at the LHC.

The simulation of different polarisation states allows us to study effects from physics beyond the Standard Model in more detail. In this project we want to study, how effective-field-theory extensions of the Standard Model will affect different polarisations of W bosons in W±W± pair production at the LHC.

The complexity of matrix element calculations for final states with a high multiplicity is so high, that their computational cost often limits theory simulations for the LHC. In this project we want to extend a machine learning based approach to simplify these calculations by approximating the matrix elements. The goal is to study various parameters for the architecture and training of deep neural networks in this approach.

For Heavy Ion collisions the Sherpa simulation program has to be able to simulate not only collisions from proton beams but also from neutron beams. This can be done by using some of the (few) available neutron PDFs or by a simple implementation of the (approximate) strong isospin symmetry, using proton PDFs with a swap of u and d quark densities. In this Bachelor project such a swapping should be implemented and validated against existing neutron PDF sets.

We will study a special effect in the matrix element of W+gamma production, where the differential cross section vanishes due to one specific polarisation combination. To this end we will use the Sherpa simulation program and analyse the simulation in analogy to analyses done with the ATLAS detector.

tbd

More information to follow.

Bachelor Thesis, Research Studies Master, Master Thesis in Experimental Particle Physics

• Multivariate analysis, machine learning and artificial intelligence
• Optimisation of data analyses for particle searches and reconstruction of particle decays
• Application and development of software for statistical data analysis

One of the research activities of the ATLAS group at the Institute of Nuclear and Particle Physics is the search for new Higgs bosons in extensions of the Standard Model at the Large Hadron Collider (LHC). A large data set is available from ATLAS Run-2 and Run-3 with a total luminosity of 200 fb^-1, which can be analysed. More data are currently being recorded by ATLAS.

In many scenarios beyond the Standard Model, the decay of the new Higgs bosons is into a pair of tau leptons. We therefore study the hadronic decay of such tau leptons in detail, in particular, to obtain a high significance for the Higgs signal, and a large suppression of backgrounds.

In the data analyses, we exploit advanced statistical methods and very often apply methods of machine learning to optimize the selection algorithms or parameter settings.

The topics of the Bachelor and Master theses will be defined individually according to your interest and to the current state of research.

In the thesis project, you will learn software-based methods of data analysis and statistics, the application of machine learning tools, and the use of modern particle detectors.

You should bring a basic knowledge of particle physics. Programming skills would be an asset, but we will also provide introductory sessions to all necessary programming and analysis tools.

© CERN

Bachelor Thesis, Research Studies Master, Master Thesis in Experimental Particle Physics

• Machine learning und artificial neural networks
• Optimisation of the energy measurement of photons, electrons and hadronic jets with Liquid-Argon calorimeters
• Programming or simulation of electronic and digital signal processing
• Statistical analysis of measurement data

The Liquid-Argon Calorimeters (LAr Calorimeters) of the ATLAS detector at the LHC are going to be upgraded with new readout electronics for operating at highest LHC luminosities. The calorimeter signals shall be reconstructed with improved energy and spatial resolution so that particles produced in proton-proton collisions can be identified with higher precision. The goal is, for example, an improved detection of Higgs boson decays with the ATLAS detector.

The particle identification and energy reconstruction must be performed in real-time, and the time to provide the energy calculation in each detector cell must not take longer than 0.5 μs. For this reason, we use modern and fast programmable electronic circuits, so-called Field Programmable Gate Arrays (FPGAs) for signal reconstruction. We exploit deep learning methods and artificial intelligence to optimize our measurements.

In the Bachelor or Master thesis, the energy reconstruction of the ATLAS LAr Calorimeters shall be improved using machine learning approaches.

In the research project, you will learn the application of machine learning tools (e.g. keras), software-based data analysis, and the functioning of a modern particle detector and electronic readout systems.

If you are interested you may also learn about programming of FPGAs using the programming language VHDL. You can get involved in the optimisation and application of machine learning algorithms running on FPGA devices.

You should bring a basic knowledge of particle physics, and the motivation to learn more about machine learning and/or FPGA programming. Basic programming skills are an asset. We also provide introductory sessions to VHDL and the analysis tools needed for your work.

© CERN

© IKTP