Speaker
Description
Accurate air-shower core reconstruction is an important basis for determining the geometry of extensive air showers, improving the reconstruction accuracy of related physical quantities, and reducing systematic uncertainties in cosmic-ray composition and energy-spectrum measurements. LHAASO-KM2A and WFCTA observe extensive air showers from different aspects: KM2A records the spatial distribution and arrival-time information of secondary particles at the ground, while WFCTA provides complementary information on shower development and geometry by measuring Cherenkov-light images. The combined use of multi-detector observations is expected to make fuller use of the multiscale responses of air showers at the ground and in the atmosphere, thereby improving the accuracy of shower-core reconstruction.
In recent years, deep-learning methods have demonstrated considerable potential for applications in high-energy physics and cosmic-ray experiments. Compared with conventional methods, neural networks can learn complex nonlinear mapping relationships from detector observations, providing a new approach for processing multi-detector, nonlinear, and multidimensional information.
Using iron simulation events generated with the CORSIKA software and the EPOS-FLUKA hadronic interaction model, we developed a ParticleNet-based graph neural network for core reconstruction in the zenith-angle range of 0°–20° and the energy range of 100 TeV–1 PeV. First, a KM2A-only model was developed using KM2A data, and its performance was superior to that of the conventional reconstruction method over the full energy range. On this basis, WFCTA information was further incorporated to construct a joint KM2A–WFCTA model. The results show that the performance of the joint model was further improved, with the core resolution improving by approximately 6%–24% over the full energy range compared with the conventional method.
This study demonstrates that multi-detector information fusion based on graph neural networks can effectively improve the core reconstruction performance, providing a technical basis for improving the accuracy of LHAASO cosmic-ray composition and energy-spectrum measurements, as well as for developing joint multi-detector geometry reconstruction methods.
