Bayesian inference constraints on jet quenching across centrality, beam energy, and observable classes in LHC heavy-ion collisions

Abstract: Jet quenching in heavy-ion collisions probes parton energy loss in the quark--gluon plasma (QGP), but the extracted transport properties may not be universally constrained across centrality, beam energy, and observable class. In this work, we perform an analysis of the compatibility and predictive transferability of Bayesian constraints obtained from a six-parameter JETSCAPE effective energy-loss model across these subsets. The model is calibrated to charged-hadron and inclusive-jet data from ALICE, ATLAS, and CMS in PbPb collisions at $\sqrt{s_{\mathrm{NN}}}=5.02$ and $2.76$ TeV. We find that centrality-dependent posteriors are largely compatible, whereas beam-energy and observable-class splits exhibit moderate shifts within overlapping credible regions, indicating that posterior overlap alone does not guarantee predictive universality. This is further examined by propagating subset posteriors to complementary datasets without refitting, where predictive performance varies across subsets. These results indicate that different observables probe distinct aspects of jet--medium interactions and motivate leading-hadron-selected jet observables to bridge hadron-biased and jet-inclusive constraints.

• The paper demonstrates that Bayesian inference can extract and constrain the jet transport coefficient using subset-restricted calibrations of high-pT observables.
• It employs the JETSCAPE model with Gaussian process emulators and MCMC sampling to validate parameter sensitivity and cross-predictive performance.
• Results reveal observable and beam-energy dependent shifts, underscoring the need for targeted cross-validation in extracting QGP transport properties.

Bayesian Inference Constraints on Jet Quenching Across Centrality, Beam Energy, and Observable Classes in LHC Heavy-Ion Collisions

Introduction and Motivation

This paper presents a comprehensive Bayesian analysis of jet quenching in PbPb collisions at $\sqrt{s_\mathrm{NN}}=5.02$ and $2.76$ TeV at the LHC, focusing on the extraction and universality of the jet transport coefficient $\hat{q}$ using the JETSCAPE effective energy-loss model. The authors address the longstanding issue of whether a unified set of jet-quenching parameters can quantitatively describe observables across centrality, collision energy, and observable class—particularly the high-$p_T$ charged-hadron nuclear modification factor $R_\mathrm{AA}$ and inclusive-jet $R_\mathrm{AA}$—and if so, under which conditions this universality is maintained.

Unlike previous global fits that collapse all high-$p_T$ jet-quenching data into a single posterior, this study performs subset-restricted calibrations, evaluating both the compatibility (posterior overlap in parameter space) and predictive transferability (accuracy when posteriors from one subset are propagated to different observables without refitting). This approach rigorously tests the operational universality of $\hat{q}$ and related energy-loss parameters.

Model Framework and Bayesian Calibration

The theoretical framework leverages the JETSCAPE multistage event generator, parameterized by a six-dimensional vector $(Q_0, \tau_0, A, B, C, \alpha_s)$ controlling the virtuality dependence, onset time, and normalization of the transport coefficient. The QCD medium is modeled via event-by-event TRENTo + 2+1D MUSIC hydrodynamics, with consistent initial-state and medium parameters across energies to isolate the impact of hard sector calibration. The extracted observables are matched to experimental measurements from ALICE, ATLAS, and CMS, focusing on high-$p_T$ charged-hadron and $2.76$0 anti-$2.76$1 jet $2.76$2 in well-specified centrality bins.

Bayesian inference is performed through Gaussian process emulators trained on a Latin hypercube sample of model parameters, with Markov-chain Monte Carlo used for sampling the posterior. Both experimental and emulator-induced covariances are propagated in the likelihood, and extensive validation, including closure tests and emulator performance assessments, confirm the statistical robustness of the inference pipeline.

Results: Parameter Constraints and Subset Decompositions

The global (all-observable) posterior exhibits strong contraction for $2.76$3 and $2.76$4, reflecting dominant experimental sensitivity to the overall quenching scale. Posterior distributions for subleading parameters ($2.76$5) remain weakly constrained, indicating limited current sensitivity to the precise functional form of $2.76$6 beyond its normalization and reference-scale value.

Critically, the authors perform systematic subset-restricted calibrations along three axes: centrality (central vs. mid-central collisions), beam energy ($2.76$7 TeV vs. $2.76$8 TeV), and observable class (charged hadron vs. inclusive jet). The following major trends are observed:

• Centrality: The $2.76$9 posteriors for central and mid-central selections exhibit substantial overlap, with minimal median shift ($\hat{q}$0), suggesting compatibility in parameter space.
• Beam Energy: Posteriors display a measurable shift; the 5.02 TeV calibration favors higher $\hat{q}$1 than 2.76 TeV. The distributions overlap but are not coincident ($\hat{q}$2).
• Observable Class: The jet-only calibration yields a broader and higher $\hat{q}$3 posterior compared to hadron-only, yet with significant overlap ($\hat{q}$4).

These findings signal that, while a single effective description broadly accommodates all data, meaningful residual dependencies exist, especially between collision energies and observable classes.

Cross-Prediction and Transfer Performance

Cross-prediction tests—which propagate posteriors inferred from one subset to predict excluded observables without retraining—demonstrate that posterior overlap alone is not a sufficient condition for predictive universality. Notably:

• Central $\hat{q}$5 Mid-central: The central calibration underpredicts mid-central data in a significant fraction of bins, with lower predictive overlap.
• Beam-Energy Transitions: Transferring posteriors between 5.02 and 2.76 TeV leads to persistent observable-level discrepancies in both directions, reflecting sensitivity to initial-state or medium/parton-spectra differences.
• Hadron-to-Jet/Jet-to-Hadron: Transfers between hadron and jet observable classes remain comparatively stable, despite their distinct sensitivities (e.g., surface-bias for hadrons vs. full in-cone shower energy for jets).

These tests highlight the necessity of explicit observable-level validation when assessing the universality of energy-loss frameworks.

Sensitivity Analysis and Interpretation

A parameter sensitivity map around the MAP point reveals leading constraints on $\hat{q}$6 and $\hat{q}$7, affirming that current high-$\hat{q}$8 inclusive suppression data predominantly constrain the normalization and reference scale of $\hat{q}$9, while remaining largely insensitive to finer details of the transport coefficient's functional form. Jet and hadron observables probe different regions of phase space—jets being less surface-biased and more susceptible to out-of-cone energy transport, with hadron observables biased towards hard-fragmenting, quark-like partons.

Implications and Outlook

The results establish that while current LHC inclusive $p_T$0 data can be described within a unified JETSCAPE-based energy-loss framework, observable- and system-dependent shifts in the extracted $p_T$1 remain at the nontrivial level. Practically, this implies that future extraction of QGP transport properties should not rely solely on global fits; subset and cross-predictive consistency checks are essential to expose latent non-universalities.

Theoretically, the observed tension across beam energy and observable class axes suggests unresolved scale dependencies or missing physics (e.g., differences in recoil, medium response, or shower evolution) in the effective model. As the authors note, differential observables that interpolate between hadron and jet constraints, such as leading-hadron-selected jets, offer a promising avenue for disentangling these effects.

Further progress will require both the inclusion of more granular and differentiated experimental inputs, and a more flexible theoretical parameterization of jet--medium dynamics, coupled to systematic Bayesian inference at the highest available level of experimental precision.

Conclusion

Through a rigorous Bayesian analysis, this paper demonstrates both the power and the limitations of current jet quenching models in describing high-$p_T$2 suppression across centrality, beam energy, and observable class at the LHC. The study provides a quantitative foundation for next-generation energy-loss extractions and highlights the necessity of cross-predictive, subset-resolved validation when pursuing universal QGP transport coefficients in jet quenching phenomenology (2604.15737).