First search for sterile neutrino oscillation leading to $\nu_{\mu}$ disappearance in the Booster Neutrino Beam at ICARUS

This paper presents the first sterile neutrino oscillation search by the ICARUS detector using the Booster Neutrino Beam, finding no statistically significant muon neutrino disappearance at the 600-meter baseline and setting 90% C.L. exclusion contours, while highlighting that the result is currently systematics-limited and will be strengthened by future combined analyses with the SBND detector.

The Gist

Imagine the universe is filled with tiny, ghost-like particles called neutrinos. They are so shy and light that they can pass through entire planets without bumping into anything. Scientists have long suspected that there might be a "fourth type" of these ghosts, called sterile neutrinos, that are even shyer and don't interact with anything at all. If they exist, they would cause the three known types of neutrinos to "disappear" or change into this invisible fourth type as they travel.

This paper is a report from a team of scientists using a giant, high-tech camera called ICARUS to hunt for these disappearing ghosts. Here is the story of their search, explained simply.

1. The Setup: A Neutrino Factory and a Giant Camera

Think of Fermilab (a particle accelerator in Illinois) as a massive factory that shoots a beam of muon neutrinos (one of the three known types) at a target. This beam is like a high-powered hose spraying invisible water.

The ICARUS detector is a 760-ton tank filled with ultra-pure liquid argon, located 600 meters (about four football fields) away from the source. It acts like a giant, 3D camera. When a neutrino finally decides to bump into an argon atom inside the tank, it creates a spark of light and a trail of electric charge, leaving a "fingerprint" that the camera records.

2. The Mystery: The "Vanishing Act"

The scientists have a specific theory to test: The 3+1 Model.

• The 3: The three known neutrino flavors (electron, muon, tau).
• The +1: The hypothetical "sterile" neutrino.

If the sterile neutrino exists, some of the muon neutrinos traveling from the factory to the camera should vanish into this invisible fourth state. The scientists expected to see fewer neutrinos arriving at the camera than the factory predicted.

3. The Hunt: Two Different Ways to Look

To make sure they didn't miss anything, the team used two different "software detectives" to analyze the data:

• Pandora: A traditional, rule-based detective that looks for patterns like a human would, carefully connecting dots to build a picture.
• SPINE: A modern, AI-powered detective (using Machine Learning) that learns to recognize patterns by studying thousands of simulated examples, much like how a child learns to recognize a cat by seeing many pictures of cats.

They looked for a very specific "fingerprint" in the tank: a muon (a heavy cousin of an electron) and at least one proton (a piece of the atom) created by the neutrino collision. They called this the 1µNp event. It's like looking for a specific type of footprint in the mud.

4. The Challenge: Noise and Fog

The biggest problem wasn't finding the neutrinos; it was knowing exactly how many should be there in the first place.

• The Fog (Systematic Uncertainties): The scientists had to guess how many neutrinos the factory was shooting and how they would behave when hitting the argon. Their "guess" had a lot of fog around it (uncertainty).
• The Analogy: Imagine trying to count how many raindrops hit your roof. If you don't know exactly how hard the storm is blowing (the flux) or how much water the roof absorbs (the interaction), it's hard to say if you're missing drops because of a leak (oscillation) or just because your math is fuzzy.

In this experiment, the "fog" was so thick that it was hard to tell if a missing neutrino was a ghost or just a math error.

5. The Result: No Ghosts Found (Yet)

After analyzing data collected in 2022 and 2023, the scientists compared what they saw in the tank to what their simulations predicted.

• The Verdict: The number of neutrinos they found matched the prediction perfectly. There was no "missing" neutrinos.
• The Conclusion: They found no evidence that muon neutrinos are turning into sterile neutrinos. The "ghosts" aren't hiding in this specific spot.

6. What's Next? The Two-Camera Strategy

The paper admits that this single-camera search was limited by the "fog" of uncertainty. To clear the fog, they have a plan:

• They are building a second, smaller camera called SBND right next to the neutrino factory (only 110 meters away).
• The Strategy: By comparing the "near" camera (which sees the beam before it has a chance to vanish) with the "far" camera (ICARUS), they can cancel out the fog. It's like having a reference photo of the rain at the source to perfectly calibrate your roof counter.

Summary

The ICARUS team took a huge, careful look at a beam of neutrinos to see if any were turning into invisible "sterile" ghosts. Using two different computer brains to analyze the data, they found that no neutrinos disappeared. While this doesn't prove sterile neutrinos don't exist, it rules out a large chunk of the possibilities. The real power of the experiment is just beginning, as they prepare to combine data from two detectors to clear the fog and get a much sharper view of the universe's secrets.

Technical Summary

1. Problem Statement

The paper addresses the long-standing anomalies in neutrino physics, specifically the LSND and MiniBooNE experiments which suggested the existence of neutrino oscillations at a short baseline ($L/E \sim 1$ km/GeV). These anomalies imply the existence of a fourth, "sterile" neutrino state with a mass-squared splitting $\Delta m^2_{41} \sim 1$ eV$^2$, which does not interact via the weak force. While previous experiments have placed limits on this hypothesis, definitive confirmation or exclusion requires high-precision measurements with controlled systematic uncertainties.

The specific challenge addressed here is the first search for muon neutrino ($\nu_\mu$) disappearance using the ICARUS detector, located 600 meters downstream of the Fermilab Booster Neutrino Beam (BNB). Unlike previous single-detector searches, this analysis aims to establish a baseline for the Short-Baseline Neutrino (SBN) program, which will eventually combine data from near (SBND) and far (ICARUS) detectors to constrain systematic uncertainties on flux and interaction models.

2. Methodology

Experimental Setup

• Beam: The Booster Neutrino Beam (BNB) at Fermilab, producing a $\nu_\mu$-dominated beam (approx. 93%) with an average kinetic energy of $\sim 0.8$ GeV.
• Detector: ICARUS, a Liquid Argon Time Projection Chamber (LArTPC) containing 760 tons of liquid argon. It is located 600 meters from the target.
• Data: Analysis of data collected during ICARUS Run 2 (Winter 2022 – Spring 2023), corresponding to an exposure of $\sim 1.6 \times 10^{20}$ Protons on Target (POT).

Event Selection (Signal Definition)

The analysis targets Charged-Current (CC) interactions of $\nu_\mu$ on argon nuclei resulting in a $1\mu Np$ final state (one muon and at least one proton).

• Criteria:
  • Exactly one contained muon track with length $> 50$ cm.
  • At least one proton with deposited energy $> 50$ MeV.
  • No additional charged pions or electromagnetic showers.
  • All particles contained within the fiducial volume to allow momentum reconstruction via range.
• Background Rejection: Utilization of the Cosmic Ray Tagger (CRT) and scintillation light (PMTs) to reject out-of-time cosmic rays. The $1\mu Np$ topology naturally suppresses Neutral Current (NC) backgrounds.

Reconstruction Frameworks

To ensure robustness and evaluate systematic uncertainties, the analysis employs two independent reconstruction algorithms:

1. Pandora: A traditional pattern recognition software using a multi-algorithm approach (Boosted Decision Trees, topological clustering) to reconstruct 3D tracks and showers from 2D wire signals.
2. SPINE (Scalable Particle Imaging with Neural Embeddings): A Machine Learning-based, end-to-end reconstruction chain using Convolutional Neural Networks (CNNs) and Graph Neural Networks (GNNs) to process voxelized 3D data.

Simulation and Systematics

• Simulation: Neutrino interactions are simulated using GENIE (AR23 model) with the LArSoft framework. Cosmic rays are modeled with CORSIKA.
• Systematic Uncertainties: The analysis is systematics-dominated. Uncertainties are categorized into:
  • Flux: Beam focusing, hadron production ($\pi, K$), and secondary interactions.
  • Interaction: Quasielastic (QE) form factors (RPA, $z$-expansion), Meson Exchange Currents (MEC), Resonance production, and Final State Interactions (FSI).
  • Detector: Electron lifetime, recombination models, TPC noise, and spatial non-uniformity.
• Fitting Framework: A new framework, PROfit, was developed for the SBN program. It uses a combined Neyman–Pearson $\chi^2$ metric, treating systematics via covariance matrices and nuisance parameters (pull terms).

Statistical Analysis

• Model: A 3+1 sterile neutrino model approximated as a two-flavor oscillation for $\nu_\mu$ disappearance.
• Hypothesis Testing: The observed spectrum is compared to the expectation under the null hypothesis (no oscillation) and the sterile neutrino hypothesis.
• Confidence Intervals: Constructed using the Feldman-Cousins procedure to ensure correct frequentist coverage, particularly near physical boundaries.
• Blinding: The analysis followed a strict blind procedure, with selection criteria and fitting tools developed on 10% of data and simulation only, unblinding only after the analysis strategy was frozen.

3. Key Contributions

1. First ICARUS Oscillation Analysis: This is the inaugural oscillation search using ICARUS data from the BNB, validating the detector's performance and reconstruction tools for the SBN program.
2. Dual-Reconstruction Approach: The simultaneous use of Pandora (traditional) and SPINE (ML-based) provides a unique cross-check. While the results are not fully independent (sharing simulation and calibration), the differing algorithms help evaluate reconstruction-related systematic uncertainties.
3. PROfit Framework: Introduction and application of the PROfit fitting framework, which handles complex systematic correlations and nuisance parameters specifically for SBN analyses.
4. Comprehensive Systematics Evaluation: A detailed breakdown of flux, interaction, and detector uncertainties, demonstrating that the analysis is currently limited by the lack of a near detector to constrain the flux and interaction models in situ.

4. Results

• Observation: No statistically significant evidence for $\nu_\mu$ disappearance was observed.
• Goodness of Fit:
  • Pandora: $\chi^2/ndof = 15.3/17$, $p$-value = 0.910.
  • SPINE: $\chi^2/ndof = 17.3/17$, $p$-value = 0.421.
  • Both analyses are consistent with the null hypothesis (no oscillation).
• Exclusion Contours: The paper presents 90% Confidence Level (C.L.) exclusion contours in the $\Delta m^2_{41}$ vs. $\sin^2 2\theta_{\mu\mu}$ plane.
  • The data contours fall within the expected sensitivity bands (Asimov sensitivity).
  • The exclusion power is strongest in regions where oscillations would manifest as a normalization deficit, which is constrained by the data-MC agreement.
• Systematic Pulls: The fit showed significant pulls on the CC-MEC normalization (pulled to $+1\sigma$) and RPA suppression parameters. This indicates that the data prefers a slightly higher event rate than the central simulation, which is accommodated by these interaction model uncertainties rather than by oscillation parameters.

5. Significance and Future Prospects

• Consistency: The results are consistent with other single-detector disappearance searches (MINOS+, NOvA, CDHS) and the global picture of short-baseline anomalies, though they do not resolve the tension with the IceCube closed contour.
• Systematics Limitation: The analysis confirms that single-detector searches are limited by large a priori uncertainties in neutrino flux and interaction modeling (total systematic uncertainty $\sim 20\%$).
• Path Forward: The primary significance of this work is as a stepping stone. The SBN program's ultimate goal is a two-detector analysis combining ICARUS (far) and SBND (near, 110m).
  • The near detector will provide in situ measurements of the unoscillated flux and interaction cross-sections.
  • This will constrain the systematic uncertainties by orders of magnitude, enabling a world-leading, robust test of the 3+1 sterile neutrino hypothesis with $5\sigma$ sensitivity.

In summary, this paper establishes the ICARUS detector's capability to perform precision oscillation physics, validates the SBN analysis tools, and sets the stage for the definitive two-detector sterile neutrino search in the near future.