A flow-matching generative model for event-by-event jet-induced hydro response in high-energy heavy-ion collisions

Kai-Yi Wu, Zhong Yang, Long-Gang Pang, Xin-Nian Wang · 2026

Read A flow-matching generative model for event-by-event jet-induced hydro response in high-energy heavy-ion collisions by Kai-Yi Wu, Zhong Yang, Long-Gang Pang, Xin-Nian Wang (2026) 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-2605-17511.txt.

Introduction

Wu and colleagues train a flow-matching generative model on coupled transport-hydrodynamic events to study jet-induced medium response in heavy-ion collisions. Given only initial photon and jet spatial and momentum information, the model conditionally reproduces final-state hadron spectra from gamma-jet-induced response in central lead-lead collisions. It runs about a million times faster than full simulations. It preserves the Mach-cone front, diffusion wake and their spectral contributions, plus wake-driven rapidity asymmetry.

Catalog record

CREATORKai-Yi Wu, Zhong Yang, Long-Gang Pang, Xin-Nian Wang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58046
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.17511

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A flow-matching generative model for event-by-event jet-induced hydro response in high-energy heavy-ion collisions Kai-Yi Wu, Zhong Yang, Long-Gang Pang, Xin-Nian Wang Abstract In high-energy heavy-ion collisions, propagation of the energy deposited into the medium by energetic partons that traverse the quark-gluon plasma (QGP) leads to Mach-cone-like jet-induced medium response. Full simulations of such jet-induced medium responses require a complete model such as the coupled Linear Boltzmann Transport and hydrodynamic (CoLBT-hydro) model that can carry out the concurrent evolution of both hard partons and the medium. Such full simulations on parallelized computers, however, are very resource-intensive and alternative simulation methods will be useful for more extensive physics investigations. In this study, we train a Flow Matching generative model with γ \gamma -jet events in 0-10 % \% Pb+Pb collisions at s NN \sqrt{s_{\rm{NN}}} = 5.02 TeV from the CoLBT-hydro model to estimate the final-state hadron spectra d 3 ​ N / d ​ p T ​ d ​ η ​ d ​ ϕ d^{3}N/dp_{T}d\eta d\phi from jet-induced hydro response. With only the initial spatial and momentum information of the γ \gamma and jets, the network is shown to conditionally generate the marginal final-state hadron spectra from the jet-induced hydro response that agree well with the training data. This generative model achieves a computational acceleration of approximately six orders of magnitude compared to the full CoLBT-hydro simulations, while faithfully preserving the statistical properties of the front and diffusion wake of the Mach-cone-like hydro response and their contributions to the hadron spectra. I Introduction Jets are collimated hadronic flows produced through the fragmentation of high-energy partons (quark or gluon) in high-energy collisions, manifesting highly concentrated clusters of hadrons in the detector. In ultra-relativistic heavy-ion collisions, partons from the initial hard scatterings interact with the quark-gluon plasma (QGP) and lose energy during their propagation in the hot medium, leading to the suppression of high transverse momentum ( p_{T} ) hadrons and jets, often referred to as jet quenching 15 ; 39 ; 53 ; 24 ; 95 ; 17 ; 77 ; 22 ; 21 ; 91 . The phenomenon can serve as a crucial probe of parton dynamics and properties of the QGP medium in high-energy heavy-ion collisions. The suppression of high- p_{T} hadron and jet cross section in high-energy heavy-ion collisions has been observed in experiments at RHIC 10 ; 11 ; 38 ; 94 and LHC 3 ; 7 ; 25 ; 62 ; 67 , indicating the formation of quark-gluon plasma (QGP) in high-energy heavy-ion collisions. The effective parton and jet energy loss extracted from the experimental data is quite large and depends on the initial jet energy and system size 46 ; 96 ; 14 . The energy and momentum lost by jets are deposited into the QGP medium and induce a medium response in the form of Mach-cone-like excitation 23 ; 87 ; 78 ; 36 ; 76 ; 57 which in turn can modify the internal structure of the jet 30 ; 102 ; 101 ; 103 ; 104 ; 26 , affecting many observables, such as jet R_{AA} 44 ; 93 ; 2 ; 80 ; 56 ; 82 ; 9 ; 64 ; 69 , jet shape 79 ; 89 ; 59 ; 100 ; 102 ; 81 and jet fragmentation functions 92 ; 28 ; 30 ; 83 ; 4 ; 31 ; 1 . The interaction involving large-momentum-transfer between hard partons and the QGP medium can be described using perturbative Quantum Chromodynamics (pQCD) 37 ; 108 ; 63 . However, pQCD proves inadequate for describing interaction between soft recoil and thermal medium partons with small momentum transfer which are responsible for the jet-induced medium response. Therefore, jet-induced medium response is typically modeled using the relativistic hydrodynamics with a source term describing the energy and momentum deposition by a propagating jet parton. This approach can accurately capture the dynamical evolution of the medium perturbed by jets and the deposited energy and momentum. However, it does not consider the effect of the perturbed medium on the subsequent propagation of jet partons. A complete description of jet-medium interactions in high-energy heavy-ion collisions necessitates an integrated framework capable of simultaneously describing both hard and soft degrees of freedom. The coupled Linear Boltzmann Transport and hydrodynamic (CoLBT-hydro) model 28 ; 29 ; 109 based on this idea combines the Linear Boltzmann Transport (LBT) model 60 ; 45 and event-by-event (3+1)D CCNU-LBNL viscous hydrodynamic (CLVisc) model 73 ; 74 ; 70 ; 97 ; 98 ; 99 to simultaneously evolve both hard jet and recoil partons and the bulk medium with hydrodynamic response caused by the energy-momentum deposited into the medium by jet partons. This model has successfully described both jet suppression and modification of jet substructures, such as jet shape, fragmentation functions and jet-hadron correlations. The predicted signals of the diffusion wake 30 ; 102 ; 106 ; 107 ; 105 , a small depletion of particles in the opposite direction of jets, have recently been observed by experiments 26 ; 6 ; 42 at LHC. Simulations of these jet events in heavy-ion collisions using CoLBT-hydro model were carried out in parallelized GPU computers and are very resource-intensive. In order to make the simulations more accessible for more extensive studies, it is essential to develop alternative models that are fast and yet can faithfully describe the coupled evolution of both the jet and medium evolution, including the jet-induced medium response. We will develop an AI-based model for this purpose in this study, using data generated from the CoLBT-hydro model. AI for Science has rapidly emerged as a transformative paradigm, driving significant advancements at the frontiers of science research. Generally, machine learning methods can be categorized into discriminative and generative models. Generative models, in particular, are designed to learn the underlying statistical distributions of training data, enabling the efficient sampling of new data points. Prominent generative frameworks include Variational Autoencoders (VAEs) 52 ; 33 , Generative Adversarial Networks (GANs) 35 , Transformers 90 , Flow-based Models 32 , Flow Matching 58 , and Diffusion Models 85 ; 48 ; 86 . Each architecture offers distinct advantages: VAEs are known for their explicit latent space structure. GANs excel at generating high-fidelity samples through adversarial training. And Flow-based models (including Flow Matching) provide a mathematically grounded approach to modeling distributions via invertible transformations or vector fields. Unlike GANs, which rely on adversarial training, and diffusion models, which typically generate samples by iteratively reversing a stochastic noise-addition process, Flow Matching learns a continuous vector field that transports samples from a simple prior distribution to the data distribution, often along a straight-line interpolation path. Samples are then generated by solving the corresponding ordinary differential equation, leading to more stable and efficient training while maintaining high sample quality. In the field of high-energy nuclear physics, generative models have been extensively applied across various stages of heavy-ion collisions to accelerate Monte Carlo simulations 72 ; 61 ; 112 ; 43 ; 18 ; 111 ; 71 ; 47 . In Lattice QCD, flow-based models serve as efficient samplers for field configurations, which are crucial for determining the Equation of State (EoS) 12 ; 41 ; 113 ; 8 . Generative models have successfully accelerated simulations

Read the complete text