Feasibility of event-by-event primary mass discrimination using radio observables and supervised machine learning

TL;DR

The paper investigates the feasibility of event-by-event primary-mass discrimination for ultra-high-energy cosmic rays using radio observables alone, without reconstructing the shower maximum. It employs a random-forest classifier trained on per-antenna peak electric-field amplitudes and spectral slopes derived from a fast radio-emission model (RDSim) fed by ZHAireS inputs, with EM-energy normalization and conservative energy-smearing. Discrimination accuracies of 81%–96% are reported across zenith angles, demonstrating that radio observables carry sufficient composition information in principle, particularly when combining amplitude and spectral-slope features. The study presents a detector-agnostic feasibility assessment, outlines limitations, and highlights pathways for extending to multi-class mass discrimination and more realistic detector scenarios.

Abstract

In this work, we investigate the feasibility of event-by-event primary mass discrimination using radio observables only. Although the analysis does not require an explicit reconstruction of the shower maximum ($X_{max}$), the discrimination power still arises from the sensitivity of the radio observables to the longitudinal development of the extensive air shower (EAS). Such radio-based approaches could be particularly relevant for radio-only experiments, such as GRAND. To assess this feasibility, we obtained conservative upper limits for the discrimination accuracy using a supervised machine-learning (ML) algorithm, namely a random forest (RF). The input features used were the peak electric fields and the spectral slopes, which have complementary discrimination power, along with the antenna distances to the shower axis. The RF was trained and tested using large event sets generated by the fast radio emission simulation and simplified detector response implemented in the RDSim framework. We obtained discrimination accuracies between 81\% and 96\% over the studied zenith range, even after normalizing each shower by its own electromagnetic energy. Since the analysis includes deliberately conservative choices, such as a large 10\% uncertainty on the reconstructed EM energy, these quoted values should be interpreted as conservative upper limits suitable for a feasibility assessment. Our results demonstrate that event-by-event primary mass discrimination using radio observables is, in principle, feasible.

Figures (9)

• Figure 1: Left: Theoretical polarization directions of the Askaryan (red) and geomagnetic (blue) emission mechanisms. $\vec{V}$ (black) denotes the shower axis and $\vec{B}$ (magenta) the geomagnetic field. Right: Schematic of the elliptical symmetry at ground level used by the superposition emission model. The reference line, where full simulations are performed, is shown in red, while the antenna position for which the electric field is estimated is shown in blue. Both panels are adapted from toymodel.
• Figure 2: Comparison between full ZHAireS simulations and the results of the superposition model for the peak electric field amplitudes. Shown are the amplitudes for an $80^\circ$ shower along the major axis of the elliptical radio footprint. The curves labeled "With Scaling" ("No Scaling") correspond to results with (without) the early-late correction. Including this correction brings the superposition model into very good agreement with the full simulation, with a maximum difference of about 6% in this example.
• Figure 3: Example comparison between full ZHAireS simulations and the results of the emission model for the spectral slope of the radio emission. Shown are the spectral slopes for a $66^\circ$ shower along the EW (left) and NS (right) directions. In both cases, the emission model shows very good agreement with the full simulation. The blue arrow indicates the position of the reference line.
• Figure 4: RF discrimination accuracy vs zenith angle. See text for more details.
• Figure 5: Top: Feature importances at $\theta=82^\circ$, normalized to the most important feature, shown as a function of the average antenna distance to the shower axis. Overlaid are the spectral slopes obtained from the full ZHAireS simulations at the same zenith. Bottom left: Same as top, but for $\theta=62^\circ$ and with amplitudes overlaid instead. Bottom right: Same as left, but with spectral slopes overlaid.