Mass Composition Sensitivity of an Array of Water Cherenkov and Scintillation Detectors

We consider a hybrid array composed of scintillation and water Cherenkov detectors designed to measure the cosmic ray primary mass composition at energies of about 1 EeV. We have developed a simulation and reconstruction chain to study the theoretical performance of such an array. In this work we investigate the sensitivity of mass composition observables in relation to the geometry of the array. The detectors are arranged in a triangular grid with fixed 750 m spacing and the configuration of the scintillator detectors is optimized for mass composition sensitivity. We show that the performance for composition determination can be compared favorably to that of Xmax measurements after the difference in duty cycles is considered.

💡 Research Summary

The paper presents a detailed simulation study of a hybrid surface detector array designed to measure the mass composition of ultra‑high‑energy cosmic rays (UHECRs) around 1 EeV. The proposed array consists of a conventional water‑Cherenkov detector (WCD) grid with a fixed 750 m triangular spacing, overlaid with a second grid of plastic scintillator stations. Each scintillator station is built from 3 cm thick scintillator tiles (similar to those used in the KASCADE muon detectors) and can be equipped with a lead photon converter of variable thickness (0–2 radiation lengths). The authors explore a wide parameter space: primary particle type (proton, iron), primary energy (10^17.75, 10^18, 10^18.25 eV), zenith angle (0°, 25°, 38°, 48°, 60°), scintillator area (3 m², 10 m²), muon scaling factor (nominal, ×2 to mimic the muon excess observed in data), and lead shielding thickness. In total 600 configurations are simulated, each with 100 showers, yielding 60 000 events generated with CORSIKA (QGSJET II) and processed through a Geant4‑based detector response model within the Pierre Auger offline framework.

The core observable pair is the integrated signal from the scintillator array (S_sci) and the signal from the WCD at a reference distance (S_WCD). Both signals are expressed in logarithmic form and modeled as linear functions of log E and log A (mass number): log S_i = a_i + b_i log E + c_i log A. By fitting these relations to the simulated data, the authors obtain unbiased estimators for the primary energy and the logarithmic mass, effectively achieving a simultaneous reconstruction of both quantities from a single event. The discrimination power between proton‑induced and iron‑induced showers is quantified using a merit factor f = (h_s F_e – h_s p_i q) / √(σ²_F_e + σ²_p), where h_s denotes the separation of the signal means and σ the corresponding standard deviations.

A key finding is that the lateral distribution functions (LDFs) of the two detector types intersect at a distance that depends on energy and zenith angle. For the energies and angles considered, the optimal distance for the scintillator signal is around 400 m from the shower core, while the WCD uses the conventional 450 m reference (the distance where the signal variance is minimal for a 750 m grid). At small core distances (<200 m) the scintillator signal is dominated by the electromagnetic component, whereas the WCD already contains a substantial muon contribution. This complementary behavior maximizes the proton‑iron separation near the core.

The merit factor analysis shows that larger scintillator stations (10 m²) provide a higher separation power than the smaller (3 m²) stations, and that adding a thin lead converter (≈0.25 X₀) yields only marginal improvement. Scaling the muon component by a factor of two—intended to emulate the muon excess reported by the Pierre Auger Observatory—does not significantly degrade the composition sensitivity, indicating robustness against model uncertainties.

To assess practical composition determination, the authors perform a simple classification exercise: they define a cut in the two‑dimensional (log S_sci, log S_WCD) space that separates proton‑like from iron‑like events, then estimate the proton fraction f in a mixed sample using the observed counts above and below the cut. They validate the method with k‑fold cross‑validation and compare the resulting statistical uncertainty to that obtained from a traditional X_max measurement using fluorescence detectors. After accounting for the much lower duty cycle of fluorescence telescopes (≈10 % versus ≈100 % for surface detectors), the hybrid scintillator/WCD array achieves comparable uncertainties in the proton fraction, demonstrating that a surface‑only system can rival fluorescence‑based composition studies.

In summary, the simulation demonstrates that a hybrid array with 10 m² scintillator stations, placed on a 750 m triangular grid and read out at 400 m from the core, can provide an unbiased energy estimator and a mass estimator with discrimination power comparable to X_max techniques, even when realistic muon excesses and detector limitations are included. The addition of photon converters does not substantially improve performance, suggesting that resources are better allocated to increasing scintillator area or reducing grid spacing. Future work should explore tighter station spacing and dynamic optimization of the reference distance as a function of energy and zenith angle, which could further enhance composition sensitivity and help resolve the long‑standing questions about the origin and acceleration mechanisms of ultra‑high‑energy cosmic rays.