Niklas Götz

SPARKX is an open-source Python package developed to analyze simulation data from heavy-ion collision experiments. By offering a comprehensive suite of tools, SPARKX simplifies data analysis workflows, supports multiple formats such as OSCAR2013, and integrates seamlessly with SMASH and JETSCAPE/X-SCAPE. This paper describes SPARKX’s architecture, features, and applications and demonstrates its effectiveness through detailed examples and performance benchmarks. SPARKX enhances productivity and precision in relativistic kinematics studies.

This thesis investigates transport coefficients in relativistic heavy-ion collisions using a hybrid approach. Relativistic heavy-ion collisions recreate extreme conditions of temperature and density, enabling the study of the quark-gluon plasma and the QCD phase diagram. To describe the full dynamical evolution, the SMASH-vHLLE hybrid model is employed, combining hadronic transport for the non-equilibrium initial and late stages with viscous hydrodynamics for the dense intermediate stage. A key focus is the determination of shear and bulk viscosities, η/s and ζ/s, as functions of temperature and baryochemical potential. The sensitivity of final-state observables to different initial state models (IP-Glasma, TRENTo, SMASH) is analyzed, demonstrating that initial transverse momentum distributions strongly influence flow anisotropies. To constrain transport coefficients, Bayesian inference is applied, combining experimental data with model predictions in a statistically consistent framework. Results indicate a strong temperature dependence of shear viscosity, vanishing near the crossover temperature, and substantial bulk viscosity values. The baryon-density dependence of shear viscosity remains less constrained, highlighting the need for more experimental data at finite μB. This work represents the first Bayesian extraction of temperature- and density-dependent viscosities in hybrid models with hadronic initial states. The findings improve our understanding of QCD matter under extreme conditions and have implications for future experiments at FAIR and astrophysical studies of neutron stars.

Background: Significant theoretical uncertainties exist with respect to the initial condition of the hydrodynamic description of ultrarelativistic heavy-ion collisions. Several approaches exist, of which some contain initial momentum information. Its impact is commonly assumed to be small and final flow is seen as a linear response to the initial-state eccentricity.

DOI: https://doi.org/10.1103/PhysRevC.109.049901

Many high-energy-physics (HEP) simulations for the LHC rely on Monte Carlo using importance sampling by means of the VEGAS algorithm. However, complex high-precision calculations have become a challenge for the standard toolbox, as this approach suffers from poor performance in complex cases. As a result, there has been keen interest in HEP for modern machine learning to power adaptive sampling. While previous studies have shown the potential of normalizing-flow-powered neural importance sampling (NIS) over VEGAS, there remains a gap in accessible tools tailored for non-experts. In response, we introduce ZüNIS, a fully automated NIS library designed to bridge this divide, while at the same time providing the infrastructure to customise the algorithm for dealing with challenging tasks. After a general introduction on NIS, we first show how to extend the original formulation of NIS to reuse samples over multiple gradient steps while guaranteeing a stable training, yielding a significant improvement for slow functions. Next, we introduce the structure of the library, which can be used by non-experts with minimal effort and is extensivly documented, which is crucial to become a mature tool for the wider HEP public. We present systematic benchmark results on both toy and physics examples, and stress the benefit of providing different survey strategies, which allows higher performance in challenging cases. We show that ZüNIS shows high performance on a range of problems with limited fine-tuning.

Many high-energy-physics (HEP) simulations for the LHC rely on Monte Carlo using importance sampling by means of the VEGAS algorithm. However, complex high-precision calculations have become a challenge for the standard toolbox, as this approach suffers from poor performance in complex cases. As a result, there has been keen interest in HEP for modern machine learning to power adaptive sampling. While previous studies have shown the potential of normalizing-flow-powered neural importance sampling (NIS) over VEGAS, there remains a gap in accessible tools tailored for non-experts. In response, we introduce ZüNIS, a fully automated NIS library designed to bridge this divide, while at the same time providing the infrastructure to customise the algorithm for dealing with challenging tasks. After a general introduction on NIS, we first show how to extend the original formulation of NIS to reuse samples over multiple gradient steps while guaranteeing a stable training, yielding a significant improvement for slow functions. Next, we introduce the structure of the library, which can be used by non-experts with minimal effort and is extensivly documented, which is crucial to become a mature tool for the wider HEP public. We present systematic benchmark results on both toy and physics examples, and stress the benefit of providing different survey strategies, which allows higher performance in challenging cases. We show that ZüNIS shows high performance on a range of problems with limited fine-tuning.