Sensitivity to Cabibbo-Suppressed Λ Production in MicroBooNE ^†

Abstract

The MicroBooNE detector is a liquid argon time projection chamber (LArTPC) with an 85 ton active mass that receives flux from the Booster Neutrino and the Nutrinos from the Main Injector (NuMI) beams, providing excellent spatial resolution of the reconstructed final-state particles. Since 2015, MicroBooNE has accumulated many neutrino and anti-neutrino scattering events with argon nuclei enabling searches for rare interaction channels. The Cabibbo-suppressed production of hyperons in anti-neutrino–nucleus interactions provides sensitivity to a range of effects, including second-class currents, SU(3) symmetry violations and reinteractions between the hyperon and the nuclear remnant. This channel exclusively involves anti-neutrinos, offering an unambiguous constraint on wrong-sign contamination. The effects of nucleon structure and final state interactions are distinct from those affecting the quasielastic channel and modify the $\Lambda$ and $\Sigma$ production cross sections in different ways, providing new information that could help to break their degeneracy. Few measurements of this channel have been made, primarily in older experiments such as Gargamelle. We present the sensitivity of the MicroBooNE experiment to the cross section for direct (Cabibbo-suppressed) $\Lambda$ production in muon anti-neutrino interactions, using anti-neutrinos from the off-axis NuMI beam.

1. Introduction

The Cabibbo-suppressed production of hyperons is a rare neutrino interaction process with few measurements [1,2,3,4,5,6], subject to unique nuclear effects such as hyperon-nucleus potentials which modify the cross sections of the individual channels in different ways. This creates potential to disentangle the physics of the initial anti-neutrino–nucleon interaction from final-state interactions [7]. This paper presents the sensitivity of the MicroBooNE experiment [8] to the cross section of Cabibbo-suppressed $\Lambda$ production:

$$\begin{array}{c}\hfill {\overline{\nu }}_{\mu }+\mathrm{Ar}\to {\mu }^{+}+\Lambda +X.\end{array}$$

(1)

This calculation assumes data corresponding to $2.2\times {10}^{20}$ protons on target (POT) of neutrino mode flux and $4.9\times {10}^{20}$ POT of anti-neutrino mode flux from the Neutrinos from the Main Injector (NuMI) beam [9,10], which will be analyzed. The data from these periods are combined to minimize the statistical uncertainty in the measurement.

2. Event Selection

Neutrino interactions are reconstructed using the Pandora [11] reconstruction/pattern recognition framework, which identifies a candidate neutrino interaction vertex and reconstructs the resulting tracks and electromagnetic showers. These objects are analyzed by event selection, which applies a set of criteria to reject background interactions. Prior to applying any event selection, 37 signal events are predicted, among a background of 1.6 million events. The background comprises other neutrino interactions and cosmic rays mistakenly tagged as particles resulting from neutrino interactions.

Any events in which the reconstructed vertex is outside of the fiducial volume used in Ref. [12] are rejected. The selection searches for the $\Lambda \to p+{\pi }^{-}$ decay, which creates a V-shaped signature in the detector, seen in Figure 1. There are three final-state particles that must be identified: the ${\mu }^{+}$, p and ${\pi }^{-}$, all of which produce track-like signatures. Any events containing fewer than three reconstructed tracks or any showers are rejected. Particle identification (PID) scores [13] are employed to analyze the calorimetric information associated with these tracks, indicating if they are muon-like or proton-like. The longest muon-like track is the muon candidate. A pair of tracks consistent with a proton and charged pion are identified using an array of boosted decision trees [14], leveraging variables such as PID scores [13] and Pandora’s track/shower classification score [15]. From this pair of tracks, two kinematic quantities are calculated: their invariant mass, displayed in Figure 2a, and angular deviation, denoted as $\alpha$ in Figure 2b. The angular deviation is defined as the angle between the reconstructed momentum vector of the $\Lambda$ candidate and the line joining the primary vertex to the decay vertex. Cuts are applied to these variables to remove events with kinematics inconsistent with a decaying $\Lambda$ baryon. Finally, the “island finding” method, illustrated in Figure 3, is employed to determine whether the proton and pion form a true displaced vertex.

After applying the selection, the surviving events consist of the signal, other events in which $\Lambda$ baryons are produced (such as deep inelastic scattering or resonance excitation), other hyperons such as the ${\Sigma }^0$, events containing secondary interactions of neutrons and events containing reconstruction problems. The efficiency of the event selection is $7\%$ with a final purity of $47\%$, suppressing the background by a factor of ∼10${}^6$.

3. Systematic Uncertainties

Four sources of systematic uncertainties are considered: the neutrino flux [16], cross sections of background neutrino interaction processes [17], cross sections of secondary interactions simulated by Geant 4 [18,19] and the detector response [20]. The dominant source of uncertainty in the predicted signal is the flux.

4. Cross-Section Extraction

The cross-section extraction procedure involves generating pseudo-experiments and repeatedly calculating the cross section in each to estimate the effect of uncertainties, including the asymmetric shapes of the statistical uncertainties. To propagate the MC simulation statistical uncertainties, the Bayesian posterior distributions of the selection efficiency $ϵ$, and the selected background, B, are calculated with uniform priors using the TEfficiency class [21]. The data’s statistical uncertainty is modeled by applying Bayes’ theorem to the Poisson distribution, yielding the posterior distribution $P\left(N\right|N_{\mathrm{obs}})$ when $N_{\mathrm{obs}}$ events are observed in the data.

The systematic uncertainties are propagated by first calculating the covariance matrix of the selection efficiency $ϵ$, the total muon anti-neutrino flux $\Phi$, and the selected background B and constructing a three-dimensional Gaussian distribution, from which systematic fluctuations ${\alpha }_{ϵ}, {\alpha }_{\Phi }$ and ${\alpha }_B$ may be sampled. Pseudo-experiments are then generated by sampling values of $ϵ, B, N, {\alpha }_{ϵ}, {\alpha }_{\Phi }$ and ${\alpha }_B$ from their respective distributions and calculating their resulting cross section with

$$\begin{array}{c}\hfill {\sigma }_{*}=\frac{N-(B+{\alpha }_B)}{T(\Phi +{\alpha }_{\Phi })\Gamma (ϵ+{\alpha }_{ϵ})}.\end{array}$$

(2)

The resulting distributions are drawn, for $N_{\mathrm{obs}}$ ranging from 0 to 7, in Figure 4, compared with the cross section predicted by MicroBooNE’s GENIE tune [17].

5. Summary

A complete procedure for the extraction of the cross section of the Cabibbo-suppressed production of $\Lambda$ baryons has been developed, and is ready for application to MicroBooNE data. A sophisticated event selection is employed, reducing an initial background of 1.6 million events to ∼3 events, against a predicted signal of 2.5 events. The final sensitivity is expressed as a collection of Bayesian posterior probability distributions, calculated by generating pseudo-experiments in which the quantities used to calculate the cross section are varied according to their uncertainties.