\section{Discussion}

In this thesis the yet unobserved rare \B-decay \Btokpipiee at \lhcb was studied using a total amount of $1\invfb$ of data. The same mode with \mup\mun in the final state has already been observed and could be used for ratio tests of the lepton flavour universality.

First, the stripping lines have been applied to the sample. A strong preselection is then applied in order to remove physical contributions to the background. This is necessary to further reduce the combinatorial background in our signal region with a MVA and to have an unbiased yield. The yield is estimated and a blind fit to the vertex constrained \B invariant mass is performed successfully.

It turned out that the cuts were insufficient as peaks are occurring in the data right next to the blinded region, which most probably come from physical background and reaches into our signal region. It is further to note that the classifier used in the MVA showed an unexpected high performance. This can happen if the classifier actually trains on the signature of physical background which differs much more from the signal then combinatorial background does. Another reason for this could be the differences between generated and real events which leads to the classifier being trained to distinguish those two instead of signal versus background. On one hand the generated and real sample do not seem to differ too large for this decay and on the other hand a multidimensional reweighting procedure has been applied in order to further reduce the differences.

To continue with this analysis and to perform a first detection, it would be a possible step to perform a more in-deep study of the background and to remove any remaining physical contributions from it.