Search for: All records

Abstract Scattering of high energy particles from nucleons probes their structure, as was done in the experiments that established the non-zero size of the proton using electron beams 1 . The use of charged leptons as scattering probes enables measuring the distribution of electric charges, which is encoded in the vector form factors of the nucleon 2 . Scattering weakly interacting neutrinos gives the opportunity to measure both vector and axial vector form factors of the nucleon, providing an additional, complementary probe of their structure. The nucleon transition axial form factor, F A , can be measured from neutrino scattering from free nucleons, ν μ n  →  μ − p and $${\bar{\nu }}_{\mu }p\to {\mu }^{+}n$$ ν ¯ μ p → μ + n , as a function of the negative four-momentum transfer squared ( Q 2 ). Up to now, F A ( Q 2 ) has been extracted from the bound nucleons in neutrino–deuterium scattering 3–9 , which requires uncertain nuclear corrections 10 . Here we report the first high-statistics measurement, to our knowledge, of the $${\bar{\nu }}_{\mu }\,p\to {\mu }^{+}n$$ ν ¯ μ p → μ + n cross-section from the hydrogen atom, using the plastic scintillatormore »target of the MINERvA 11 experiment, extracting F A from free proton targets and measuring the nucleon axial charge radius, r A , to be 0.73 ± 0.17 fm. The antineutrino–hydrogen scattering presented here can access the axial form factor without the need for nuclear theory corrections, and enables direct comparisons with the increasingly precise lattice quantum chromodynamics computations 12–15 . Finally, the tools developed for this analysis and the result presented are substantial advancements in our capabilities to understand the nucleon structure in the weak sector, and also help the current and future neutrino oscillation experiments 16–20 to better constrain neutrino interaction models.« less

Abstract We compare different neural network architectures for machine learning algorithms designed to identify the neutrino interaction vertex position in the MINERvA detector. The architectures developed and optimized by hand are compared with the architectures developed in an automated way using the package “Multi-node Evolutionary Neural Networks for Deep Learning” (MENNDL), developed at Oak Ridge National Laboratory. While the domain-expert hand-tuned network was the best performer, the differences were negligible and the auto-generated networks performed as well. There is always a trade-off between human, and computer resources for network optimization and this work suggests that automated optimization, assuming resources are available, provides a compelling way to save significant expert time.