<p>We present track reconstruction algorithms based on deep learning, tailored to overcome specific central challenges in the field of hadron physics. Two approaches are used: (i) deep learning (DL) model known as fully-connected neural networks (FCNs), and (ii) a geometric deep learning (GDL) model known as graph neural networks (GNNs). The models have been implemented to reconstruct signals in the non-Euclidean detector geometry of the future antiproton experiment PANDA. In particular, the GDL model shows promising results for cases where other, more conventional track-finders fall short: (i) tracks from low-momentum particles that frequently occur in hadron physics experiments and (ii) tracks from long-lived particles such as hyperons, hence originating far from the beam-target interaction point. Benchmark studies using Monte Carlo simulated data from PANDA yield an average technical reconstruction efficiency of 92.6% for high-multiplicity muon events, and 97.1% for the <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$\Lambda$</EquationSource> <EquationSource Format="MATHML"><math> <mi mathvariant="normal">Λ</mi> </math></EquationSource> </InlineEquation> daughter particles in the reaction <InlineEquation ID="IEq2"> <EquationSource Format="TEX">$\bar{p}p \rightarrow \bar{\Lambda }\Lambda \rightarrow \bar{p}\pi ^+ p\pi ^-$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mover accent="true"> <mrow> <mi>p</mi> </mrow> <mrow> <mo stretchy="false">¯</mo> </mrow> </mover> <mi>p</mi> <mo stretchy="false">→</mo> <mover accent="true"> <mrow> <mi mathvariant="normal">Λ</mi> </mrow> <mrow> <mo stretchy="false">¯</mo> </mrow> </mover> <mi mathvariant="normal">Λ</mi> <mo stretchy="false">→</mo> <mover accent="true"> <mrow> <mi>p</mi> </mrow> <mrow> <mo stretchy="false">¯</mo> </mrow> </mover> <msup> <mi>π</mi> <mo>+</mo> </msup> <mi>p</mi> <msup> <mi>π</mi> <mo>-</mo> </msup> </mrow> </math></EquationSource> </InlineEquation>. Furthermore, the technical tracking efficiency is found to be larger than 70% even for particles with transverse momenta <InlineEquation ID="IEq3"> <EquationSource Format="TEX">$p_\text {T}$</EquationSource> <EquationSource Format="MATHML"><math> <msub> <mi>p</mi> <mtext>T</mtext> </msub> </math></EquationSource> </InlineEquation> below 100 MeV/<i>c</i>. For the long-lived <InlineEquation ID="IEq4"> <EquationSource Format="TEX">$\Lambda$</EquationSource> <EquationSource Format="MATHML"><math> <mi mathvariant="normal">Λ</mi> </math></EquationSource> </InlineEquation> hyperons, the track reconstruction efficiency is fairly independent of the distance between the beam-target interaction point and the <InlineEquation ID="IEq5"> <EquationSource Format="TEX">$\Lambda$</EquationSource> <EquationSource Format="MATHML"><math> <mi mathvariant="normal">Λ</mi> </math></EquationSource> </InlineEquation> decay vertex. This underlines the potential of machine-learning-based tracking, also for experiments at low- and intermediate-beam energies.</p>

Application of Geometric Deep Learning for Tracking of Hyperons in a Straw Tube Detector

期刊论文
出版模式：

开放获取
发表日期：

2025年10月21日

查看全文

Adeel Akram,
Xiangyang Ju,
Michael Papenbrock,
Jenny Taylor,
Tobias Stockmanns,
Karin Schönning

摘要

We present track reconstruction algorithms based on deep learning, tailored to overcome specific central challenges in the field of hadron physics. Two approaches are used: (i) deep learning (DL) model known as fully-connected neural networks (FCNs), and (ii) a geometric deep learning (GDL) model known as graph neural networks (GNNs). The models have been implemented to reconstruct signals in the non-Euclidean detector geometry of the future antiproton experiment PANDA. In particular, the GDL model shows promising results for cases where other, more conventional track-finders fall short: (i) tracks from low-momentum particles that frequently occur in hadron physics experiments and (ii) tracks from long-lived particles such as hyperons, hence originating far from the beam-target interaction point. Benchmark studies using Monte Carlo simulated data from PANDA yield an average technical reconstruction efficiency of 92.6% for high-multiplicity muon events, and 97.1% for the $\Lambda$ $Λ$ daughter particles in the reaction $\bar{p}p \rightarrow \bar{\Lambda }\Lambda \rightarrow \bar{p}\pi ^+ p\pi ^-$ $\bar{p} p \to \bar{Λ} Λ \to \bar{p} π^{+} p π^{-}$ . Furthermore, the technical tracking efficiency is found to be larger than 70% even for particles with transverse momenta $p_\text {T}$ $p_{T}$ below 100 MeV/c. For the long-lived $\Lambda$ $Λ$ hyperons, the track reconstruction efficiency is fairly independent of the distance between the beam-target interaction point and the $\Lambda$ $Λ$ decay vertex. This underlines the potential of machine-learning-based tracking, also for experiments at low- and intermediate-beam energies.