NOvA

A dramatisation of the NOvA detector — Credit: Symmetry Magazine.

The NuMI Off-axis ν_e Appearance (NOvA) experiment is a long-baseline neutrino oscillation experiment near Ash River, Minnesota which has been recording data since 2014.

NOvA has two detectors, separated by 810 km. In addition to the Far Detector in Ash River, MN, there is also a Near Detector at Fermilab, IL, that measures the neutrino beam before oscillations occur.

NOvA measures the 750 kW beam of muon neutrinos produced by the NUMI beam line, and the Far Detector measures both electron neutrinos and muon neutrinos due to oscillations. Neutrino oscillations are a quantum mechanical phenomenon where neutrinos created in one flavor state are observed interacting as different flavor states after travelling a given distance.

Three Flavor Neutrino Oscillations

The contour for the 2020 NOvA Three Flavor Oscillation Analysis.

The primary research goal of NOvA is measuring three flavor neutrino oscillations. Prof. Sanchez led the effort on the first electron neutrino appearance analysis in NOvA, and previously co-lead all analysis efforts in NOvA. Group members have worked across all aspects of the analysis, and have made significant contributions to every result which has been presented.

These contributions have included ensuring the quality of data collected by the experiment, and checking that the simulation and recorded data are in agreement. Group members have also been responsible for performing corrections between the near and far detectors, as well as optimising event selection categories. The most recent NOvA oscillation results can be found here.

Some of the crucial contributions of the group to the wider physics goals of the experiment are discussed below.

Neutrino Cross Section Physics

The NOvA Collaboration Cross Section working group is currently working on a triple differential muon neutrino charged current inclusive cross section measurement, which is now in the final stages prior to publication. With this as a solid stepping stone, the ISU group is working on more exclusive cross section channels. We are developing machine learning-based tools to allow for pion rejection aiming towards a muon neutrino charged current pionless or quasielastic-like cross section measurement. We have also started studies on proton reconstruction in the NOvA Near Detector that could enable reporting cross sections on proton multiplicity or kinematics.

Measurement of Neutrons

Neutrons candidate events broken down by particle type — Neutron candidate events broken down by particle type.

Neutrons are one of the primary resultant particles of antineutrino interactions and we can see them in the NOvA detectors. This means it is vital for us to accurately model how often neutrons interact and how much energy they leave in the detectors when they do. The neutrino group has made substantial progress in proving the simulation does a poor job of modeling the neutrons that are in the detector and are striving to pinpoint the exact physical processes that are modeled incorrectly. An algorithm which uses reconstructed variables to count energy deposits linked back to neutron interactions has been developed and used to compare data and simulation. Neutrons do not deposit energy on their own, the resultant particles of their interactions do, so a neural network capable of identifying if an energy deposit came from the resultant particle of a neutron and the type of resultant particle is in development. An initial pass at a network shows promising results.

Reconstruction and Deep Learning

Separation of the different event types the NOvA event classification network is trained against.

Accurately identifying neutrino interactions is a difficult challenge, but one that must be overcome. To do this NOvA employs both traditional and deep learning techniques. Dr. Warburton, a postdoc in the group, has led these efforts since 2017. The group uses industry standard deep learning libraries, Keras and Tensorflow, and has developed a Pandas based analysis framework to perform network inference and evaluation on GPUs resources available at Fermilab.

NOvA has networks which perform both interaction classification (separating muon neutrino interactions from electron neutrino interactions) and individual particle classification (separating muons, electrons and pions). These networks were first introduced in 2016, and have been critical parts of all NOvA analyses since then; having effectively increased our recorded by 30% compared to using traditional techniques.

Joint Analysis with T2K

The Super Kamiokande detector in Japan — The Super Kamiokande Detector in Tokai, Japan

Tokai to Kamioka (T2K) is a competing long-baseline neutrino oscillation experiment in Japan. It has a shorter baseline of 295 km, studies neutrinos of lower energies and uses a different detector technology to NOvA. These differences mean that a joint analysis of the two experiments can provide much greater resolution to the parameters which govern neutrino oscillations.

For this reason Prof. Sanchez is leading the efforts to perform a joint analysis between the two experiments. This will require overcoming differences in simulation, analysis and error handling strategies. The first joint analysis is planned for 2021/22.

Three Flavor Neutrino Oscillations

Neutrino Cross Section Physics

Measurement of Neutrons

Reconstruction and Deep Learning

Joint Analysis with T2K

Notable Recent Talks and Posters

Notable References