A First Determination of the LHC Neutrino Fluxes from FASER Data

TL;DR

This paper presents a novel, machine learning-based method to measure and analyze the LHC forward neutrino fluxes using FASER data, providing new insights into particle production and potential new physics.

Contribution

It introduces NN$ u$flux, a theory-agnostic, machine learning approach for extracting LHC neutrino fluxes from experimental data, validated through closure tests and applied to real FASER data.

Findings

First extraction of LHC muon neutrino flux from FASER data

Discriminates between different forward hadron production models

Constrains BSM scenarios and proton intrinsic charm

Abstract

The detection of TeV neutrinos from the LHC by the far-forward detectors FASER and SND@LHC enables a plethora of novel physics opportunities. Among these, the measurement of the flavour, energy, and rapidity dependence of the LHC forward neutrino fluxes provides unique constraints on theoretical predictions of forward particle production in hadronic collisions. We demonstrate that neutrino event yield measurements at FASER from Run 3 and at its HL-LHC upgrades enable a theory-agnostic extraction of the LHC forward neutrino fluxes. We exploit the equivalence of the problem with the determination of parton distributions from deep-inelastic structure functions to apply the NNPDF approach, based on machine learning regression and the Monte Carlo replica method, to LHC neutrino fluxes. The resulting NN$\nu$flux methodology is validated through closure tests and applied to a first extraction of the LHC muon neutrino flux from the FASER 2024 data. We show how NN$\nu$flux can discriminate between event generators of forward hadron production; scrutinize a possible intrinsic charm component in the proton; and constrain BSM scenarios with enhanced decays of neutral hadrons into neutrinos.