WorksPhysics and Astronomy
Depth of Maximum of EAS and Mean Mass Composition of Primary CRs from the Spatial and Temporal Distribution of Cherenkov Light According to TAIGA-HiSCORE Data
Abstract
We present the dependence of the depth of maximum of extensive air showers (EAS), , on the energy of the primary particle, as determined from data recorded with the wide-angle Cherenkov array TAIGA-HiSCORE during the 2021–2025 observation seasons. The reconstruction of from the experimental data uses relations obtained from EAS simulations: the dependence on the steepness of the Cherenkov light lateral distribution function (-method) and on the full width at half maximum of the Cherenkov light pulse (-method). The reconstructed values are converted into the mean logarithmic mass , which characterizes the primary cosmic-ray mass composition. The results are compared with measurements from Tunka-133 (2017), the Pierre Auger Observatory (2019), LHAASO (2024), TALE (2026), and other experiments.
Introduction
This paper describes two methods for determining the depth of maximum, , of extensive air showers (EAS) from data of the TAIGA-HiSCORE Cherenkov array. The first, the -method (or timing method), is based on the duration of Cherenkov light pulses, measured as their full width at half maximum (FWHM, ). The second, the -method, uses the steepness of the Cherenkov light lateral distribution function (LDF).
The TAIGA-HiSCORE array
The TAIGA-HiSCORE array is part of the TAIGA Astrophysical complex, located in the Tunka Valley (Buryatia, Russia) at an altitude of 675 m above sea level [1]. It comprises 120 optical stations arranged on a square grid with a spacing of approximately 100 m, covering an area of about 1 km. Each station is equipped with four photomultiplier tubes (PMTs) and Winston cones, providing a total light collection area of 0.5 m [2]. Signals from both the anode and dynode outputs of the PMTs are recorded to extend the dynamic range of Cherenkov light measurements. The signals are digitized at intervals of 0.5 ns. During processing, signals from the different PMTs within a station can be shifted by one or more sampling intervals before summation, allowing a more accurate pulse shape to be recovered. For each recorded EAS event, the reconstruction determines the shower-core coordinates in the observation plane, the arrival direction, and the primary-particle energy [3, 4].
The array has operated in its full configuration since 2021. The stations were pointed toward the zenith during the 2022–2023 and 2024–2025 observation seasons and tilted 25° southward during the 2021–2022 and 2023–2024 seasons.
Simulations
To refine the relation between the observables described below and the distance to the shower maximum, , where is the atmospheric depth at the level of the Tunka Valley (675 g cm), EAS simulations were performed using CORSIKA [5] with the QGSJET-II-04 [6] and Sibyll 2.3d [7] hadronic interaction models. Approximately 4,000 EAS events initiated by primary protons, helium, nitrogen, and iron nuclei were generated at five energies (1, 3, 10, 30, and 100 PeV) and two zenith angles (0° and 30°).
Cherenkov light pulses were simulated with a bunch size of one photon and without statistical thinning. This increased the computational cost but improved the accuracy of the simulated pulse shapes. To account for pulse distortions introduced by the detector electronics (PMTs and preamplifiers), each simulated pulse was convolved with the instrumental response function. The mean FWHM of the instrumental response across all stations is 9.2 ns for the anode channel and 6.1 ns for the dynode channel.
The -method
The timing method (or -method [8]) relies on the calculated dependence of the width of an EAS Cherenkov light pulse on the relative position of the shower maximum, . The relationship between and the pulse duration at a fixed distance from the shower axis was obtained by fitting the simulations for four primary particle types at energies of 10–100 PeV and zenith angles up to 30°.
For a distance of 300 m from the shower axis, fitting the simulated points yields the following relationship between and for the anode channel, illustrated in Figure 1:

Figure 1. Calculated dependence of on for the anode channel. The line shows the linear fit and the shaded band shows its uncertainty.
For the dynode channel at the same distance, the corresponding relationship is
To determine the mean logarithm of the pulse duration at 300 m experimentally, stations lying 270–330 m from the axis are selected. At all simulated energies, the pulse durations at these distances exceed the instrumental response width by at least a factor of two. Moreover, the dependence of the pulse duration on distance is approximately linear, with nearly the same slope at all simulated energies. The experimental data provide an average of 5–10 measurements in the selected distance interval.
The -method
The -method [9], previously introduced to reconstruct from the steepness of the Cherenkov light LDF, has been modified using the simulation results presented in this work. The dimensionless parameter is now defined as , providing a more accurate relation to . Here, denotes the Cherenkov light flux density at a distance from the EAS axis. The values and at distances of 80 and 160 m, respectively, are obtained by interpolation of the Cherenkov light LDF fitted to the light densities measured by the detector stations.
The simulated relationship between and is
Equation (3) is independent of primary-particle energy over the range 1–30 PeV, zenith angle (), and hadronic interaction model (QGSJET-II-04 or Sibyll 2.3d). The simulated relationship is shown in Figure 2. At energies above 30 PeV, this method is limited by saturation of the measured Cherenkov pulse amplitude at distances less than 100 m from the EAS axis.

Figure 2. Calculated dependence of on the Cherenkov light LDF steepness parameter .
Analysis of experimental results
The dependence of the mean shower-maximum depth, , on primary-particle energy was determined from four TAIGA-HiSCORE observation seasons in 2021–2025 (about 900 hours of observation). Events were selected with zenith angles and shower cores within the effective detection area of approximately 1 km. The energy range was 10–250 PeV for the -method (35,453 events) and 1–30 PeV for the -method (820,854 events). The experimental values obtained with TAIGA-HiSCORE, together with results from other experiments, are shown in Figure 3.

Figure 3. Experimental dependence of the mean on the primary energy.
The mean shower-maximum depth is converted into the mean logarithmic mass using linear interpolation:
This interpolation was carried out using the QGSJET-II-04 model [6]. The numerical and values in the energy bins are listed in Table 1 and Table 2, respectively for the - and -methods.
| (g/cm) | (g/cm) | |||
| 3.045597 | 516.32 | 0.18 | 1.715 | 0.005 |
| 3.147013 | 525.98 | 0.22 | 1.633 | 0.007 |
| 3.246653 | 534.41 | 0.27 | 1.580 | 0.008 |
| 3.346587 | 542.24 | 0.33 | 1.544 | 0.010 |
| 3.446696 | 548.94 | 0.41 | 1.540 | 0.013 |
| 3.546268 | 555.80 | 0.50 | 1.529 | 0.016 |
| 3.646147 | 562.79 | 0.62 | 1.512 | 0.020 |
| 3.746112 | 568.38 | 0.78 | 1.539 | 0.026 |
| 3.845842 | 574.92 | 0.96 | 1.533 | 0.032 |
| 3.945507 | 578.51 | 1.22 | 1.623 | 0.041 |
| 4.045509 | 584.91 | 1.56 | 1.620 | 0.053 |
| 4.146435 | 588.41 | 1.96 | 1.716 | 0.067 |
| 4.246363 | 590.18 | 2.47 | 1.870 | 0.086 |
| 4.346180 | 592.63 | 3.14 | 2.003 | 0.110 |
| 4.445805 | 592.95 | 3.88 | 2.211 | 0.138 |
Table 1. The -method: TAIGA-HiSCORE data in the 1–30 PeV interval.
| (g/cm) | (g/cm) | |||
| 4.010847 | 572.78 | 0.44 | 1.955 | 0.015 |
| 4.151481 | 579.11 | 0.76 | 2.046 | 0.026 |
| 4.250379 | 581.53 | 0.92 | 2.179 | 0.032 |
| 4.351593 | 586.73 | 1.17 | 2.222 | 0.041 |
| 4.451388 | 590.17 | 1.41 | 2.322 | 0.050 |
| 4.550986 | 590.13 | 1.62 | 2.548 | 0.058 |
| 4.647211 | 595.84 | 2.17 | 2.559 | 0.079 |
| 4.741832 | 603.93 | 2.46 | 2.478 | 0.090 |
| 4.838065 | 616.06 | 3.40 | 2.247 | 0.126 |
| 4.943829 | 626.33 | 3.81 | 2.100 | 0.143 |
| 5.020429 | 636.15 | 5.81 | 1.900 | 0.219 |
| 5.118750 | 646.51 | 5.23 | 1.723 | 0.199 |
| 5.333590 | 664.14 | 9.57 | 1.511 | 0.373 |
Table 2. The -method: TAIGA-HiSCORE data for energies from 10 PeV.
The conversion results are shown in Figure 4. According to the TAIGA-HiSCORE measurements, the primary cosmic-ray composition becomes heavier between the knee region and 30–50 PeV, approaching the CNO-group value (), and then becomes lighter at higher energies, reaching at about 300 PeV. The two reconstruction methods yield shower-maximum depths differing by approximately in this energy region. For comparison, Figure 4 includes results from LHAASO [10], the Pierre Auger Observatory [11], Telescope Array [12], TALE [13], KASCADE [14], the Yakutsk EAS array [15], and Tunka-133 [4].

Figure 4. Experimental dependence of the mean on the primary energy.
The mean mass compositions derived from the charged-particle component of EAS by LHAASO (2024) and from the recently reanalysed KASCADE data by Kuznetsov and colleagues (2024) exhibit a similar energy dependence of to that obtained here.
Above 50–100 PeV, the TAIGA-HiSCORE points connect with those of the Pierre Auger Observatory, which indicate a continuing shift toward a lighter composition, approaching a purely protonic composition at 1,000 PeV. A trend toward lighter composition with increasing energy is seen in the data of the large arrays considered here.
The conversion from shower-maximum depth to mean cosmic-ray mass composition depends on the hadronic interaction model used in the EAS simulations. The present analysis primarily uses QGSJET-II-04. Models that predict more rapid dissipation of primary-particle energy at high energies place the shower maximum higher in the atmosphere. Consequently, for a primary-particle energy of 50 PeV and the same experimentally measured depth of shower maximum, the inferred increases by approximately 0.35 (22.67%) when using Sibyll 2.3d and by 0.08 (4.66%) when using EPOS-LHC, relative to QGSJET-II-04.
Conclusions
The methods for determining the EAS shower-maximum depth from TAIGA-HiSCORE Cherenkov-array measurements have been improved: the -method reconstruction algorithm was refined and a new calibration was obtained for the -method. Processing the 2021–2025 data with both methods yielded preliminary values of and in the energy range 1–250 PeV, which were compared with other experimental results. The data support a shift toward a heavier cosmic-ray mass composition from the knee to approximately 30 PeV, followed by a shift toward a lighter composition above approximately 50 PeV. The difference between the results obtained by the two -reconstruction methods is being investigated through additional detailed simulations that account for the array geometry and detector saturation close to the EAS axis. The current TAIGA-HiSCORE results should therefore be considered preliminary and subject to further revision.
Funding
The work was carried out at the MSU-ISU Astrophysical Complex (agreement EB 075-15-2021-675).
The work was supported by the Ministry of Science and Higher Education of the Russian Federation (projects FZZE-2026-0006, FZZE-2027-0007, FSUS 2025-0008, FZZE-2025-0001) and the Russian Science Foundation (grant 23-72-00016 (sections 3, 4, 5)).
Conflict of interest
The authors declare no conflict of interest.
References
- [1]N. Budney et al. Nuclear Instruments and Methods in Physics Research A, 1039:167047, 2022. doi: 10.1016/j.nima.2022.167047.
- [2]O. Gress et al. Nuclear Instruments and Methods in Physics Research A, 845:367, 2017. doi: 10.1016/j.nima.2016.08.031.DOI
- [3]V. V. Prosin et al. Bulletin of the Russian Academy of Sciences: Physics, 89(6):907–912, 2025. 10.1134/S1062873825711663.
- [4]N. M. Budney et al. Astroparticle Physics, 117:102406, 2020. doi: 10.1016/j.astropartphys.2019.102406.
- [5]Corsika software. https://www.iap.kit.edu/corsika/index.php.
- [6]S. Ostapchenko. Physical Review D, 83:014018, 2011. doi: 10.1103/PhysRevD.83.014018.DOI
- [7]F. Riehn et al. Physical Review D, 102:063002, 2020. doi: 10.1103/PhysRevD.102.063002.DOI
- [8]V. V. Prosin, S. F. Berezhnev, N. M. Budney, et al. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 756:94–101, 2014. doi: 10.1016/j.nima.2013.09.018.DOI
- [9]V. V. Prosin et al. Bulletin of the Russian Academy of Sciences: Physics, 85(4):395, 2021. doi: 10.3103/S1062873821040298.DOI
- [10]X.-J. Lv et al. The Astrophysical Journal, 979(2):225, 2025. doi: 10.3847/1538-4357/ada426.DOI
- [11]A. Yushkov and Pierre Auger Collaboration. In Proceedings of Science (ICRC2019), page 482, 2019. doi: 10.22323/1.358.0482.DOI
- [12]R. U. Abbasi et al. The Astrophysical Journal, 858(2):76, 2018. doi: 10.3847/1538-4357/aabad7.DOI
- [13]R. U. Abbasi et al. Physical Review D, 113:062003, 2026. doi: 10.1103/vrky-dxn7.DOI
- [14]M. Yu. Kuznetsov, N. A. Petrov, I. A. Plokhikh, and V. V. Sotnikov. Journal of Cosmology and Astroparticle Physics, 2024(05):125, 2024. doi: 10.1088/1475-7516/2024/05/125.DOI
- [15]S. Knurenko and I. Petrov. Advances in Space Research, 64(12):2570, 2019. doi: 10.1016/j.asr.2019.07.019.DOI