Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fast Jet Finding in Julia

Graeme Andrew Stewart. Sanmay Ganguly, Sattwamo Ghosh, Philippe Gras, Atell Krasnopolski

TL;DR

The paper addresses the need for fast, flexible jet reconstruction in collider data analysis by introducing JetReconstruction.jl, a high-performance Julia package for sequential jet finding. It details a production-ready release with support for both $pp$ and $e^+e^-$ collision algorithms, including Anti-$k_T$, Cambridge/Aachen, Inclusive-$k_T$, Durham, and Generalised $k_T$, along with SIMD-enhanced core routines and a ClusterSequence data model. Key contributions include the introduction of an $(\theta,\phi)$-space EEjet for $e^+e^-$ clustering, EDM4hep data model support, and substructure tools (Soft Drop, Mass Drop, Filtering, Trimming), backed by extensive benchmarking showing speedups over FastJet. The work enables fast, scalable jet reconstruction in Julia, with practical impact for FCCee and future colliders, and provides a pathway for broader analysis tooling and data-format interoperability within the Julia ecosystem.

Abstract

Jet reconstruction remains a critical task in the analysis of data from HEP colliders. We describe in this paper a new, highly performant, Julia package for jet reconstruction, JetReconstruction.jl, which integrates into the growing ecosystem of Julia packages for HEP. With this package users can run sequential reconstruction algorithms for jets. In particular, for LHC events, the Anti-${k}_\text{T}$, Cambridge/Aachen and Inclusive-${k}_\text{T}$ algorithms can be used. For FCCee studies the use of alternative algorithms such as the Generalised ${k}_\text{T}$ for $e^+e^-$ and Durham are also supported. The performance of the core algorithms is better than Fastjet's C++ implementation, for typical LHC and FCCee events, thanks to the Julia compiler's exploitation of single-instruction-multiple-data (SIMD), as well as ergonomic compact data layouts. The full reconstruction history is made available, allowing inclusive and exclusive jets to be retrieved. The package also provides the means to visualise the reconstruction. Substructure algorithms have been added that allow advanced analysis techniques to be employed. The package can read event data from EDM4hep files and reconstruct jets from these directly, opening the door to FCCee and other future collider studies in Julia.

Fast Jet Finding in Julia

TL;DR

The paper addresses the need for fast, flexible jet reconstruction in collider data analysis by introducing JetReconstruction.jl, a high-performance Julia package for sequential jet finding. It details a production-ready release with support for both and collision algorithms, including Anti-, Cambridge/Aachen, Inclusive-, Durham, and Generalised , along with SIMD-enhanced core routines and a ClusterSequence data model. Key contributions include the introduction of an -space EEjet for clustering, EDM4hep data model support, and substructure tools (Soft Drop, Mass Drop, Filtering, Trimming), backed by extensive benchmarking showing speedups over FastJet. The work enables fast, scalable jet reconstruction in Julia, with practical impact for FCCee and future colliders, and provides a pathway for broader analysis tooling and data-format interoperability within the Julia ecosystem.

Abstract

Jet reconstruction remains a critical task in the analysis of data from HEP colliders. We describe in this paper a new, highly performant, Julia package for jet reconstruction, JetReconstruction.jl, which integrates into the growing ecosystem of Julia packages for HEP. With this package users can run sequential reconstruction algorithms for jets. In particular, for LHC events, the Anti-, Cambridge/Aachen and Inclusive- algorithms can be used. For FCCee studies the use of alternative algorithms such as the Generalised for and Durham are also supported. The performance of the core algorithms is better than Fastjet's C++ implementation, for typical LHC and FCCee events, thanks to the Julia compiler's exploitation of single-instruction-multiple-data (SIMD), as well as ergonomic compact data layouts. The full reconstruction history is made available, allowing inclusive and exclusive jets to be retrieved. The package also provides the means to visualise the reconstruction. Substructure algorithms have been added that allow advanced analysis techniques to be employed. The package can read event data from EDM4hep files and reconstruct jets from these directly, opening the door to FCCee and other future collider studies in Julia.

Paper Structure

This paper contains 12 sections, 2 equations, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Production Release of Julia Jet Reconstruction
  3. Support for $e^+e^-$ algorithms
  4. Durham Algorithm
  5. Generalised ${k}_\text{T}$ for $e^+e^-$
  6. Implementation Details
  7. $pp$ and $e^+e^-$ Performance
  8. $pp$ Algorithms
  9. $e^+e^-$ Algorithms
  10. Substructure and Taggers
  11. FCCee Jets and EDM4hep
  12. Conclusions

Figures (4)

  • Figure 1: Visualisation of a typical $pp$ collision jet reconstruction, using anti-${k}_\text{T}$ with $R=2$, in the $y-\phi$ (rapidity, azimuthal angle) plane. The height of each bar indicates the original cluster energy and the colour represents the final jet clustering, i.e., all clusters with the same colour are clustered together.
  • Figure 2: Jet reconstruction time for anti-${k}_\text{T}$ jet reconstruction at different values of $R$, using the N2Tiled strategy. Reconstruction time for Julia and Fastjet are plotted against the average cluster density of different samples.
  • Figure 3: Jet reconstruction time for anti-${k}_\text{T}$ jet reconstruction at different values of $R$, using the N2Plain strategy. Reconstruction time for Julia and Fastjet are plotted against the average cluster density of different samples.
  • Figure 4: Jet reconstruction times for $e^+e^-$ algorithms. Reconstruction time for Julia and Fastjet are plotted against the average cluster density of different samples.