First combined tuning on transverse kinematic imbalance data with and without pion production constraints

Weijun Li, M. Roda, Júlia Tena Vidal, C. Andreopoulos, X.-g. Lu, Adi Ashkenazi, J. Barrow, S. Dytman, H. Gallagher, Alfonso Garcia, S. Gardiner, Matan Goldenberg, Robert Hatcher, O. Hen, I. Kakorin, Konstantin S. Kuzmin, Anselmo Meregalia, V. A. Naumov, A. Papadopoulou, Gabriel Perdue, Komninos-John Plows, Alon Sportes, Noah Steinberg, V. S. Syrotenko, J. Wolcott, Qiyu YanarXiv 2024 · DOI · arXiv

TL;DR

Researchers from the GENIE Collaboration have performed the first combined "tuning" of neutrino interaction models using data from two major experiments (T2K and MINERvA). By focusing on Transverse Kinematic Imbalance (TKI)—variables that are uniquely sensitive to nuclear effects—they developed a new model that successfully describes both pion-producing and pion-free scattering events for the first time.

What Problem Does This Paper Address?

Neutrinos are notoriously difficult to study because they rarely interact with matter. When they do interact, they usually hit a complex atomic nucleus (like Carbon) rather than a lone proton or neutron. This introduces "nuclear effects" that complicate our measurements:

  1. Initial State Effects: The nucleons inside the nucleus are already moving (Fermi motion) and are bound by strong forces.
  2. Final-State Interactions (FSIs): After a neutrino hits a nucleon, the resulting particles (like protons or pions) must fight their way out of the dense nucleus, often scattering or being absorbed along the way.

Current simulation tools, like the GENIE generator, have struggled to model these effects consistently across different experiments and different types of events (topologies). Specifically, models that worked for events without pions often failed miserably when pions were present, particularly for the MINERvA experiment's $\pi^0$ data. This paper aims to find a "global tune"—a set of model parameters—that works for everything at once.

Methods

The team utilized a framework called Transverse Kinematic Imbalance (TKI). TKI looks at the momentum mismatch between the outgoing lepton (muon) and the outgoing hadrons (protons/pions) in the plane transverse to the incoming neutrino beam. Because these variables are relatively independent of the neutrino's energy, they act as a "magnifying glass" for nuclear physics.

The researchers tuned two specific components of the GENIE model:

  • SF-CFG (Spectral Function-like Correlated Fermi Gas): Models the initial state of the nucleons.
  • hA (Intranuclear Cascade): A simplified, effective model for how particles scatter as they exit the nucleus.

They used the Professor tool, which reduces the computationally expensive task of running simulations to a fast polynomial interpolation, allowing them to minimize the $\chi^2$ (a measure of "mismatch") between theory and data across 14 different model parameters.

Datasets

The study used four key differential cross-section datasets, all involving neutrino-hydrocarbon (CH) scattering:

  1. T2K $0\pi$: Events with one muon and at least one proton, but no pions.
  2. T2K $\pi^+$: Events with a muon, a proton, and exactly one positive pion.
  3. MINERvA $0\pi$: Similar to T2K $0\pi$ but at higher neutrino energies.
  4. MINERvA $\pi^0$: Events with a muon, at least one proton, and at least one neutral pion.

Key Insights and Ideas

  • The $\pi^0$ Discrepancy: The default GENIE model (G24-0) significantly overestimated the number of neutral pions escaping the nucleus. The "aha moment" was realizing that the model predicted too many pions undergoing "No Interaction" and not enough being absorbed or scattered.
  • Breaking the Silo: Previously, models were tuned to individual experiments or specific topologies. This work proves that a single set of nuclear parameters can describe both the "clean" $0\pi$ events and the "messy" pion-producing events simultaneously.
  • Energy Independence: The success of the tune across both T2K (low energy) and MINERvA (high energy) validates that TKI variables are indeed robust probes of the underlying nuclear physics, rather than just artifacts of the neutrino beam energy.

Key Results

The researchers produced two new "tunes": G24-c (a reduced, stable parameter set) and G24-t (a more aggressive, comprehensive set).

  • Global Improvement: The total $\chi^2$ for the combined datasets dropped from 461.25 (default) to 385.94 (G24-t), a significant statistical improvement.
  • Fixing MINERvA: The most dramatic result was in the MINERvA $\pi^0$ channel, where the $\chi^2$ per bin dropped from roughly 70/12 to 36/12.
  • Model Parameters: The tune suggested that the fraction of nucleons in "Short-Range Correlations" (SRC) is slightly higher than previously thought (~15% vs 12%), and that neutral pion rescattering (specifically charge exchange) needs to be significantly suppressed in the models to match reality.

Limitations and Future Work

  • Effective Nature: The hA FSI model is "effective," meaning it simplifies complex multi-step scattering into a single step. While it describes the data well, it may not capture the full microscopic physics.
  • Target Material: All data used here was on hydrocarbon. Future experiments like DUNE will use Liquid Argon. The authors note that data from the MicroBooNE experiment (which uses argon) will be essential to see if these tunes hold up for heavier nuclei.
  • Degeneracy: There is still some "degeneracy" in the parameters—different combinations of initial-state and FSI settings can produce similar results. Incorporating electron-scattering data (where the "beam" energy is perfectly known) is the next logical step to break these ties.

Summary updated

Abstract

We present the first combined tuning, using GENIE, of four transverse kinematic imbalance measurements of neutrino-hydrocarbon scattering, both with and without pion final states, from the T2K and MINERvA experiments. As a proof of concept, we have simultaneously tuned the initial state and final-state interaction models (SF-CFG and hA, respectively), producing a new effective model that more accurately describes the data.
BibTeX
@misc{li2024first,
  author = {Weijun Li and M. Roda and Júlia Tena Vidal and C. Andreopoulos and X.-G. Lu and Adi Ashkenazi and J. Barrow and S. Dytman and H. Gallagher and Alfonso Garcia and S. Gardiner and Matan Goldenberg and Robert Hatcher and O. Hen and I. Kakorin and Konstantin S. Kuzmin and Anselmo Meregalia and V. A. Naumov and A. Papadopoulou and Gabriel Perdue and Komninos-John Plows and Alon Sportes and Noah Steinberg and V. S. Syrotenko and J. Wolcott and Qiyu Yan},
  title = {{First combined tuning on transverse kinematic imbalance data with and without pion production constraints}},
  year = {2024},
  eprint = {2404.08510},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2404.08510},
}
Figure