Efficient Monte Carlo Event Generation for Neutrino-Nucleus Exclusive Cross Sections
Mathias El Baz, Federico Sánchez, Natalie Jachowicz, Kajetan Niewczas, Ashish Kumar Jha, Alexis Nikolakopoulos · 2025
Read Efficient Monte Carlo Event Generation for Neutrino-Nucleus Exclusive Cross Sections by Mathias El Baz, Federico Sánchez, Natalie Jachowicz, Kajetan Niewczas, Ashish Kumar Jha, Alexis Nikolakopoulos (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-2502-14452.txt.
Introduction
Mathias El Baz and co-authors present a faster way in 2025 to simulate how neutrinos hit atomic nuclei. Precise models of these hits use heavy quantum methods, which makes them slow to run in real experiments. The team trained a flow-based model to mimic one such model from the Ghent group. The mimic produces sample events that closely track the original.
Their generator kept over ninety-eight percent of the useful sample weight, a strong accuracy mark. That means experiments can run far more events at far less cost. Neutrino studies need huge event counts to pin down particle traits. Faster generators help teams test ideas and fit data without cutting model quality.
Catalog record
| CREATOR | Mathias El Baz, Federico Sánchez, Natalie Jachowicz, Kajetan Niewczas, Ashish Kumar Jha, Alexis Nikolakopoulos |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 60 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-14452
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Efficient Monte Carlo Event Generation for Neutrino-Nucleus Exclusive Cross Sections Mathias El Baz, Federico Sánchez, Natalie Jachowicz, Kajetan Niewczas, Ashish Kumar Jha, Alexis Nikolakopoulos Abstract Modern neutrino-nucleus cross section computations need to incorporate sophisticated nuclear models to achieve greater predictive precision. However, the computational complexity of these advanced models often limits their practicality for experimental analyses. To address this challenge, we introduce a new Monte Carlo method utilizing normalizing flows to generate surrogate cross sections that closely approximate those of the original model while significantly reducing computational overhead. As a case study, we built a Monte Carlo event generator for the neutrino-nucleus cross section model developed by the Ghent group. This model employs a Hartree-Fock procedure to establish a quantum mechanical framework in which both the bound and scattering nucleon states are solutions to the mean-field nuclear potential. The surrogate cross sections generated by our method demonstrate excellent accuracy with a relative effective sample size of more than 98.4 % 98.4\% , providing a computationally efficient alternative to traditional Monte Carlo sampling methods for differential cross sections. Preprint: APS/123-QED I Introduction The study of neutrino interactions with nuclei is critical for reducing systematic uncertainties in modern neutrino oscillation experiments. Among the most important processes is Charged Current Quasi-Elastic (CCQE) scattering, where a neutrino interacts with a nucleon target inside a nucleus, emitting a charged lepton while ejecting one nucleon. This interaction forms the backbone of event reconstruction in neutrino oscillation experiments like T2K (Tokai to Kamioka experiment [ 1 ] ) and the new Hyper-Kamiokande water Cherenkov detector [ 2 ] . However, the theoretical modeling of CCQE interactions is complex due to the need to account for the nuclear environment in which the nucleons are bound originally [ 3 ] . Moreover, these experiments often use neutrino beams that span a broad energy spectrum—from a few hundred MeV to several GeV—making Monte Carlo (MC) event generators indispensable. MC generators, such as NEUT [ 4 ] , GENIE [ 5 ] , NuWro [ 6 ] and ACHILLES [ 7 ] , efficiently sample different interaction processes over many possible final states, enabling exhaustive modeling across this wide energy range. Such modeling is essential for extracting oscillation parameters and minimizing systematic uncertainties in neutrino physics. A variety of interaction models have been implemented in MC generators to describe neutrino-nucleus interactions. However, most of these models are limited to providing only inclusive cross sections and therefore cannot reliably predict final-state hadron kinematics. As a consequence, the latter are often generated in an ad hoc and unrealistic manner [ 8 , 9 ] . The only approaches currently used in neutrino generators that systematically provide final-state nucleon kinematics are based on the factorization obtained in the plane-wave impulse approximation (PWIA). In this picture, one assumes no final-state interactions, so the process can be factorized into (i) the primary neutrino-nucleon interaction and (ii) a nuclear model describing the bound nucleons. However, these approaches lack the effects of the nuclear medium on the outgoing nucleon. In the simplest instance of PWIA, the nucleus is modeled as a Relativistic Fermi Gas (RFG) [ 10 ] , in which nucleons are treated as fermions uniformly occupying momentum states up to the Fermi momentum in a constant potential. This approach has been remarkably successful in describing general properties of intermediate-energy processes, especially in inclusive electron scattering. Yet, it neglects important nuclear features such as the shell structure and nucleon-nucleon correlations. A more refined description of the nuclear medium still within the PWIA is provided by the Spectral Function (SF) model [ 11 ] , which yields a probability distribution S(p,E) for finding a nucleon with momentum p and removal energy E based on (e,e^{\prime}p) data and theoretical inputs. Although the SF model captures the spread in binding energies and some nucleon-nucleus correlations, its essential reliance on the PWIA means that correlations essential to accurately predict the final-state hadron kinematics cannot be fully taken into account. While these approximations were adequate when the focus was largely on lepton kinematics, the increased precision of upcoming accelerator-based long-baseline neutrino oscillation experiments will enable much more detailed measurements of outgoing hadron kinematics. To achieve the few-percent level of uncertainty on neutrino interactions required by these next-generation neutrino oscillation studies, cross section models must go beyond PWIA to include final-state interactions. To obtain exclusive differential cross sections, a theoretical framework that accurately describes the hadron system is required. One such approach is based on the distorted-wave impulse approximation (DWIA) in the mean field approximation [ 12 , 13 ] . Mean-field-based models describe nucleons within an averaged nuclear potential generated by all other nucleons. This naturally leads to a shell structure, where each nucleon in the initial state is described by a bound wave function. Meanwhile, the distorted-wave treatment goes beyond the plane-wave description by also considering the final-state interactions of the outgoing nucleon: instead of emerging as a free particle, the outgoing nucleon propagates through the same mean field potential and undergoes distortions due to its interactions with the residual nucleus. This approach has recently gained renewed attention in the context of event generator development, with the first implementation of a relativistic DWIA-based CCQE model in NEUT [ 14 ] . Despite their advantages, mean-field-based cross section calculations are computationally expensive compared to the factorized PWIA. The slow evaluation speed of these cross sections translates into slow sampling rates when traditional methods like accept-reject algorithms are used. To mitigate this, some current solutions [ 15 , 16 , 17 , 18 ] involve precomputing the hadron tensor components into heavy multidimensional grids, known as hadron tensor tables [ 8 , 19 , 20 , 21 ] . This method allows to compute the cross section much faster, and make an accept-reject sampling possible. However, this approach has significant drawbacks: it lacks flexibility, requiring recalculation for intricate change in theoretical parameters, and seem infeasible for multi-nucleon knockout exclusive cross sections due to the exponential scaling of table size with dimensionality. Recent advancements in Artificial Intelligence, particularly in Deep Learning, offer promising solutions to these challenges [ 22 , 23 , 24 , 25 ] . In this work, we propose an alternative sampling method based on normalizing flows (NF) [ 26 ] . Normalizing flows transform a simple base distribution, such as a Gaussian, into a complex target distribution through a series of learnable diffeomorphisms. This enables efficient sampling and evaluation of the target distribution—in this case, the neutrino-nucleus cross section. In this work, we demonstrate the capacity of NF to build a Monte Carlo event generator for neutrino-induced cross sections. For our proof-of-principle we use the Hartree-Fock mean-field exclusive 1-particle-1-hole ( 1p1h ) cross section formalism developed by the Ghent group [ 27 , 28 , 13 ] . These cross sections exhibit all the features needed to demonstrate that our approach can pave the way for faster and more flexible event generation in neutrino physics. The structure of this article is as follows: we first review the theoretical framework behind CCQE interactions within the mean-field approach adopted by the Ghent group. Following this, we introduce the normalizing flows technique and outline how it is integrated into the Monte Carlo event