AI-driven neutrino diagnostics and radiation-hard beam instrumentation for next-generation neutrino experiments
S. Ganguly, Fermi National Accelerator Laboratory, Batavia, USA · 2025
Read AI-driven neutrino diagnostics and radiation-hard beam instrumentation for next-generation neutrino experiments by S. Ganguly, Fermi National Accelerator Laboratory, Batavia, USA (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-06645.txt.
Introduction
S Ganguly proposes AI-driven diagnostics and tougher instruments for the next generation of neutrino experiments, in a 2025 paper. A new facility at Fermilab will fire an intense neutrino beam toward a giant underground detector. The science goals include precision tests of neutrino behavior and searches for charge-parity violation. Steady, well-measured beams are vital for all of them.
At the plan's heart sits a digital twin of the beam line. It rebuilds particle paths from muon profiles, spill by spill. That allows fast beam correction and a stable flux. Radiation-hard sensors feed the twin trustworthy data. The framework aims to speed the experiment's physics reach. Robust operations matter as much as raw beam power.
Catalog record
| CREATOR | S. Ganguly, Fermi National Accelerator Laboratory, Batavia, USA |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 23 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-06645
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
AI-driven neutrino diagnostics and radiation-hard beam instrumentation for next-generation neutrino experiments Thanks: S. Ganguly, Fermi National Accelerator Laboratory, Batavia, USA Abstract The Long Baseline Neutrino Facility (LBNF) at Fermilab will deliver a high-intensity, multi-megawatt neutrino beam to the Deep Underground Neutrino Experiment (DUNE), enabling precision tests of the three-neutrino paradigm, CP violation searches, neutrino mass ordering determination, and supernova neutrino studies. In order to accelerate DUNE’s physics reach and ensure robust beam operations, we propose an integrated AI-driven framework with real-time diagnostics and radiation-hardened instrumentation. A physics-informed digital twin is at the heart of this Real-Time Beam Integrity Monitor. By reconstructing pion phase space from muon profiles and exploiting magnetic horn optic linearity, it enables spill-by-spill beam correction and flux stabilization. By using this approach, flux-related systematics could be reduced from 5% to 1%, potentially accelerating the discovery of CP violations by four to six years. Complementing this, a US–Japan R&D effort will deploy a LGAD-based muon monitor in the NuMI beamline. Time of Flight (ToF) measurements can be acquired with picosecond precision using this radiation-hard system, enhancing sensitivity to horn chromatic effects. Simulations confirm strong responses to these effects. Machine learning models can predict beam quality and horn current with sub-percent accuracy. This scalable, AI-enabled strategy improves beam fidelity and reduces systematics, transforming high-power accelerator operations. 1 Introduction The Long Baseline Neutrino Facility (LBNF) [ 1 ] at Fermilab will deliver a multi-megawatt, high-intensity neutrino beam to the massive detectors of the Deep Underground Neutrino Experiment (DUNE) [ 2 ] , enabling precise tests of the three-neutrino flavor paradigm, direct searches for CP violation, determination of neutrino mass ordering, and detailed studies of supernova neutrino interactions. Although DUNE is baselined to achieve its physics goals independently, recent advancements in AI-driven diagnostics and real-time accelerator control provide a unique opportunity to surpass baseline performance at a significantly lower cost. A major driver of sensitivity in DUNE is the precision of the neutrino flux prediction. Achieving percent-level accuracy in flux normalization can reduce the overall systematics to below 3%, potentially shortening the runtime needed for key measurements by several years. Current beam flux uncertainties at the source are approximately 2-3%, and new approaches are required to suppress this contribution further. Muon monitors, placed downstream of the decay pipe, offer a unique window into the phase space of parent pions. Since muons and neutrinos are produced together in pion decays, measurements from the muon monitors provide indirect but valuable information about the resulting neutrino flux. The current implementations, however, are limited in time resolution and cannot capture correlations between events. A machine-learning-based diagnostic framework aims to improve flux prediction and enable adaptive beam control in this work. The approach uses correlations between real-time signals from muon monitors and upstream instrumentation to infer the pion phase space and chromatic beamline effects [ 3 ] , such as those arising from horn current fluctuations. These models are trained using a physics-informed digital twin, allowing us to bridge simulation and real-world conditions. Figure 1 demonstrates the impact of horn current uncertainty on neutrino flux prediction. A typical uncertainty of \pm 3 kA (1%) leads to flux variations exceeding 5% in the peak energy region, while reducing the uncertainty to \pm 0.15 kA suppresses this effect to below 1%. This motivates the need for precision monitoring and inference frameworks that can resolve such variations on a spill-by-spill basis. The machine learning strategy developed in this work aims to achieve this resolution using only muon monitor signals, enabling both tighter flux constraints and future feedback-based beam optimization. Figure 1: Relative change in neutrino flux as a function of energy for horn current variations of \pm 3 , \pm 0.5 , and \pm 0.15 kA. Smaller uncertainty in horn current reduces the fractional flux variation by up to a factor of five. 2 Muon Monitoring and Beam Correlation As part of the Neutrino Beam Instrumentation system for LBNF, dedicated detectors monitor secondary and tertiary particles. This provides essential information for beam alignment, anomaly detection, and long-term diagnostics. Its two most important components are the Hadron Alignment Detector System (HaDeS) and Muon Monitor System (MUMS). They serve as the main interface between upstream beamline parameters and downstream neutrino flux behavior. The spatial configuration of these systems is illustrated in Fig. 2 , showing the placement of muon monitors within shielded alcoves behind the hadron absorber. Figure 2: Cross-sectional schematic of the LBNF absorber and downstream monitoring region. Muon monitors are placed in three shielded alcoves downstream of the Hadron Absorber. These locations offer energy discrimination via natural pion decay kinematics. HaDeS is derived from the NuMI hadron monitor design [ 4 ] , which utilizes a 7\times 7 array of parallel-plate ionization chambers, covering a total area of 1 m 2 . For LBNF, HaDeS will be upgraded with finer spatial granularity and a higher channel count to improve resolution. During low-intensity alignment scans, it can be inserted into the beamline to assess residual hadrons after the target, horn, and baffle components are installed. With this technology, beam-based alignments and failure diagnoses are possible with high precision. The Muon Monitor System (MuMS), located downstream of the Hadron Absorber in three shielded alcoves, correlates proton beam properties with neutrino production. Each MuMS station consists of a 9\times 9 array of parallel-plate ionization chambers. Each station provides 2D charge maps for each spill, revealing the transverse distribution of muons. These distributions encode valuable information about pion focusing, decay kinematics, and the effectiveness of the horn current settings. Because pions and muons share the same parentage with neutrinos, monitoring the muon beam allows inference of fluctuations in the resulting neutrino flux. To study these relationships, we use the g4lbnf simulation package, which models the full LBNF beamline geometry. Simulations employ a Gaussian beam profile with \sigma=0.27 cm and track pion decays and resulting muons through the beamline. In the nominal geometry, Alcove 1 receives muons from lower-energy pions emitted at wide angles, while Alcoves 2 and 3 capture muons more forward-going and higher-energy. This natural energy separation across the MuMS stations provides a handle for diagnosing beam focusing and energy spectra shifts. Figure 3: Simulated muon flux distributions at each MuMS alcove from g4lbnf . Alcove 1 captures the broadest angular range. Muon flux and centroid shifts provide indirect evidence of pion focusing and target degradation. Figure 3 illustrates the simulated muon flux distributions at each MuMS alcove. The design of the MuMS enables the study of both beam position sensitivity and beam spot size dependence. Tighter beam profiles result in more localized muon distributions, making centroid shifts more sensitive to upstream beam perturbations. Moreover, the energy-selective positioning of the alcoves provides an opportunity to decouple overlapping beam effects. Future multi-MW upgrades to LBNF will require muon monitors that are both radiation-hard and capable of resolving time-structured beam features. Candidate technologies include optical fiber-based Cherenkov detectors and fast ionization chambers with sub-nanosecond-scale timing. These could provide complementary time-resolved muon flux profiles, enabling more granular ML model inputs and real-time monitoring at unprecedented beam powers. Table 1: Summary of LBNF Beamline Instrumentation and Their ML Roles Instrument Measurement Location ML Role BPM / SEM (x,y) centroid, RMS width Transport line to target Provides