Experimental Flavor Physics

Breadcrumb Navigation


Bachelor Thesis Topics


Measurement of the branching fraction of B0 -> D(*)- K+ (topic not available any more)

The theoretical calculation of decays like B0 -> D(*)- K+ requires a good understanding of hadronic effects. Recent calculations have achieved a high precision and show astonishing deviations from the measured values. In this project, a measurement of the branchig fractions of B0 -> D- K+ and B0 -> D*- K+ should be performed for the first time with data of the Belle II experiment. An independent result is an important step towards a solution of the puzzle.

Measurement of BR(Y(4S) -> B0 anti-B0) with exclusive B decays

The branching fraction of Y(4S) -> B0 anti-B0, called f00, is an important quantity for many anlasyses of B mesons. Several measurements of f00 are based on the the assumption of isospin symmetry in B meson decays, meaning it is assumed that the branching fractions of B0 and corresponding B+ decays are identical. So far there is only one measurement that does not make this assumption, but determines f00 via the rate of events where both B mesons decay to a certain final state. To achieve a high reconstruction efficiency, the final state B0 -> D*+ (-> D0 π+) l nu is used where only the pion and lepton are explicitly reconstructed. A disadvantage of this method is the high background and the resulting large systematic uncertainty.

With the increasing Belle II data volume, this problem can be addressed by exclusively reconstructing the D0 in a selected decay mode. The sensitivity of such an analysis as function of data size should be estimated.

Machine learning for an improved background suppression for a R(D*) measurement (topic not available any more)

Measurements of R(D*) = BR(B -> D* tau nu) / BR(B -> D* l nu) currently receive a lot of attention because the world average value deviates from the standard model prediction. In an ongoing analysis it is checked if after the reconstruction of the signal decay B -> D* l nu all remaining particles are compatible with originating from the second B meson in Y(4S) -> B anti-B events. This inclusive tagging method has the advantage of a higher efficiency compared with an exclusive reconstruction of decays of the second B meson, but leads also to a higher background level. Various methods of machine learning should be studied with the aim to improve the background suppression.

Application of autoencoder algorithms in fully-inclusive studies at Belle II experiments (topic not available any more)

One of the strengths of the Belle II experiment is a possibility of conducting a fully inclusive studies, in which there are no restriction being put on a hadron final state of a certain quark. Machine learning (ML) classification methods are frequently applied in this class of analyses, however, because of the specific structure of the input dataset, the number of input variables to ML algorithms can be quite large (> 70). These high-dimensional datasets can be hard to validate on a real experimental data, hence one has to apply methods to reduce their dimensionality. To reduce the dimensionality an unsupervised autoencoder neural network algorithm has been developed, which can efficiently compress a dataset with large number of dimensions to a few dimensions without a significant loss of information. One of the main goals of this thesis is to train the autoencoder algorithm on the real experimental data and compare the results to the training on the simulated data, which should give more control over the associated systematic uncertainties. Prior knowledge of machine learning techniques and python programming is advisable.

Tagging of B Mesons

A key feature of B factories is that they produce clean events of B meson pairs in reactions e+e- -> Y(4S) -> B anti-B. This feature is exploited by many analyses, e.g. for the reconstruction of B meson decays to invisible particles. In these analyses the second B meson is reconstructed in a hadronic or semileptonic decay, called tagging. The aim of this topic is the implementation of the tagging algorithm used by the BaBar experiment in the Belle II software and the comparison with the current algorithm. The BaBar algorithm uses a semi-inclusive approach where a seed meson, e.g. a D*, is reconstructed and then longlived particles are added until a suitable B candidate is formed.

Detector simulation with a Generative Adversarial Network (topic not available any more)

The simulation of the detector response to particles, that are created in collision events, is an essential method in experimental particle physics to connect measurements with theoretical calculations. The effort for a simulation can be enourmous and become a limiting factor. This is in particular the case for the simulation of those signals in the Pixel Vertex Detector (PXD) of the Belle II experiment that are generated by background. To address this problem, Generative Adversarial Networks (GANs) were developed to generate background data for entire PXD sensors. As an alternative approach, a two step procedure should be developed and studied where first the position of clusters and then the local cluster shape are generated.

Lab course intelligent background simulation (topic not available any more)

Searches for rare decays require a good understanding of backgrounds. Often huge amounts of costly produced simulated data are used for this purpose. One aspect that make this very inefficient is that events which are not selected as signal candidates still run through the full simulation chain because the selection quantities are only known afterwards.

In a lab course it should be predicted which events fullfil the selection criteria before the expensive detector simulation and event reconstruction are performed. Simulated decay chains of Y(4S) mesons at the Belle II experiment are used that can be described for example by graph NNs. The lab course will be offered for the first time in the summer semester. The material of the course should be critically checked and improved.

Matter-antimatter asymmetry as school project (topic not available any more)

An essential condition for the evolution of the universe towards the one we observe today, including ourselves, is a different behavior of matter and antimatter. Such an asymmetry was observed experimentally and Kobayashi and Maskawa who explained it theoretically were awarded the Nobel prize in 2008. Based on a lab course at the University of Melbourne, a project for a measurement of a matter-antimatter asymmetry was developed. The project should be further developed so that is is well understandable and manageable for school students and teachers.

Sustainability of the Belle II experiment

As in many other areas, sustainability becomes more and more important also in basic research. For the Belle II experiment in particular the operation of the accelerator, the detector, and the computing infrastructure are relevant. In addition, business trips of the researchers contribute to the ecological footprint. The available data on the sustainability of the Belle II experiment should be gathered and new data collected and assessed.



A structured program is offered for the bachelor theses in the summer semester.