Constraints on cosmic-ray propagation models from a global Bayesian analysis

Research in many areas of modern physics such as, e.g., indirect searches for dark matter and particle acceleration in SNR shocks, rely heavily on studies of cosmic rays (CRs) and associated diffuse emissions (radio, microwave, X-rays, gamma rays). While very detailed numerical models of CR propagation exist, a quantitative statistical analysis of such models has been so far hampered by the large computational effort that those models require. Although statistical analyses have been carried out before using semi-analytical models (where the computation is much faster), the evaluation of the results obtained from such models is difficult, as they necessarily suffer from many simplifying assumptions, The main objective of this paper is to present a working method for a full Bayesian parameter estimation for a numerical CR propagation model. For this study, we use the GALPROP code, the most advanced of its kind, that uses astrophysical information, nuclear and particle data as input to self-consistently predict CRs, gamma rays, synchrotron and other observables. We demonstrate that a full Bayesian analysis is possible using nested sampling and Markov Chain Monte Carlo methods (implemented in the SuperBayeS code) despite the heavy computational demands of a numerical propagation code. The best-fit values of parameters found in this analysis are in agreement with previous, significantly simpler, studies also based on GALPROP.

💡 Research Summary

The paper presents a comprehensive Bayesian framework for estimating the parameters of a numerical cosmic‑ray (CR) propagation model, specifically the widely used GALPROP code. While GALPROP can self‑consistently predict a broad range of observables—primary and secondary CR spectra, diffuse gamma‑ray emission, synchrotron radiation, etc.—its computational intensity has traditionally precluded full statistical inference. The authors overcome this limitation by coupling GALPROP with the SuperBayeS package, which implements both Nested Sampling and Markov Chain Monte Carlo (MCMC) algorithms.

A set of six to eight key propagation parameters is defined: the diffusion coefficient normalization D0, the diffusion spectral index δ, the convective wind speed Vc, the Alfvén speed vA governing re‑acceleration, and source spectral parameters (index and normalization) for primary nuclei. Physically motivated priors are assigned, reflecting theoretical expectations such as a Kolmogorov‑type turbulence spectrum (δ≈1/3) and previous experimental constraints.

The likelihood function incorporates five high‑precision data sets: the boron‑to‑carbon (B/C) ratio, the beryllium isotopic ratio (10Be/9Be), and the proton and helium spectra measured by AMS‑02 and PAMELA, together with ACE/CRIS data for low‑energy secondary/primary ratios. Both statistical and systematic uncertainties are treated explicitly.

To make the sampling tractable, the authors pre‑compute GALPROP outputs on a multidimensional grid spanning the prior ranges and employ high‑order interpolation to evaluate model predictions at arbitrary points. This surrogate‑model approach reduces the effective runtime by a factor of ~30 without sacrificing accuracy, allowing the Nested Sampler to efficiently estimate the Bayesian evidence and locate the global optimum, while the MCMC chains map the detailed posterior distributions.

The Bayesian evidence comparison shows that the simplest diffusion‑convection model (with optional re‑acceleration) provides the best balance between fit quality and model complexity; adding extra degrees of freedom such as spatially varying diffusion does not increase the evidence appreciably. The posterior means and credible intervals are: δ = 0.33 ± 0.04, D0 = (4.3 ± 0.5) × 10^28 cm² s⁻¹ at a reference rigidity of 4 GV, Vc ≈ 5 km s⁻¹ (with large uncertainty), and vA ≈ 30 km s⁻¹. These values are consistent with earlier semi‑analytic studies but now come with rigorously quantified uncertainties. The inferred δ supports a Kolmogorov turbulence cascade, while the non‑zero Alfvén speed is required to reproduce the low‑energy B/C data, confirming the necessity of re‑acceleration in the Galaxy.

Beyond parameter estimation, the work demonstrates the power of Bayesian model selection in CR physics. By penalizing unnecessary complexity, the framework naturally guards against over‑fitting and yields physically interpretable results. The authors argue that this methodology opens the door to robust, statistically sound investigations of topics that rely on CR propagation modeling, such as indirect dark‑matter searches, the interpretation of the diffuse gamma‑ray background, and the study of synchrotron emission in the radio to microwave bands. In summary, the paper establishes that a full Bayesian analysis of a state‑of‑the‑art numerical CR propagation code is feasible and yields results that are both scientifically reliable and computationally efficient.