Cross-section measurement of neutrino interactions in the TPC gas of the T2K off-axis near detector

The TPCs of the T2K off-axis near detector contain in total about $10{m^3}$ of an argon-based gas mixture (95\% [91\%] argon by volume [mass]) at atmospheric pressure. They have been exposed to the T2K neutrino beam for the full duration of T2K operation. It is thus possible to investigate neutrino interactions that happen inside the gas volume of the TPCs, using it as active target. This work presents the results obtained from T2K runs two through four. The data corresponds to $5.73 \times 10^{20}$ protons on target,$\sim 15\%$ of the total muon-neutrino beam exposure expected by the end of the currently approved T2K programme. Current neutrino cross-section models predict about $600$ charged-current muon-neutrino interaction events in the TPCs' fiducial volume in this data set. Neutrino events in the TPCs are selected using the new TPC reconstruction software \TREx. The selection works in two stages: First, the highest momentum, negatively charged particle that starts inside one of the TPCs is selected. This defines the vertex candidate. Then, additional cuts are applied to that candidate to reduce the out-of-fiducial-volume background. The selection reaches a purity of $\sim 26{\%}$ and an efficiency of $\sim 32{\%}$.This selection is then used to perform a measurement of the charged-current muon-neutrino cross-section in the TPC gas. The response-matrix centred approach is used. This method is based on the detector response matrix, which translates expectation values for event counts in truth space, i.e. describing the true properties of the events, into expectation values in reco space, i.e. describing the events in reconstructed variables. In it contained are the information about detector efficiency and detector smearing. Systematic detector uncertainties are also handled with the response matrix. A set of response matrices is generated according to the uncertainties in detector properties. Each matrix corresponds to one possible true detector performance and yields its own reco-space prediction for a given truth-space prediction. The different predictions can then be compared to the recorded data, and combined into a marginal likelihood. If the construction of the response matrix is done with care (ensuring its physics model independence),it can be used for tests of arbitrary theories against the data very easily. Since the set of response matrices contains the knowledge about the systematic detector uncertainties, the systematics do not have to be re-evaluated for each model that is tested. If the data and response matrix are published together, the model tests can also be done by theorists and model builders that are not intimately familiar with the detector that recorded the data. All they have to do is vary the truth space expectation values according to their models and multiply them with the matrix. Compared with the \enquote{classic} approach, where new models usually require a new dedicated analysis within the experiments, this makes tests of new model ideas much easier. A simple template fit with the nominal Neut and Genie models was done using this method. It yields a cross section per nucleon and scaled with average neutrino energy of $(0.42 \pm 0.15) \times 10^{-38}{cm^2/GeV}$ and $(1.02 \pm 0.24) \times 10^{-38}{cm^2/GeV}$ respectively