Figure 1

Schematic representation of the Cherenkov emission from a cosmic-ray primary particle and the light distribution on the ground and in the camera plane of an IACT.Reuse & Permissions

Figure 2

Simulated intensity distribution on the ground for the EAS-light and DC-light of an individual 50 TeV iron nucleus, as a function of distance from the shower core, for two different first interaction heights (the shower core is defined as the intersection point of the shower axis on the ground). The zenith angle is 0°. The drop in DC-intensity at $100/120\mathrm{m}$ reflects the first interaction height. The low intensity tail at larger radii is caused by Cherenkov light from fragments of the primary nucleus.Reuse & Permissions

Figure 3

Total number of emitted DC-photons as a function of energy and first interaction height for an iron nucleus at a zenith angle of 0°, calculated using an atmospheric profile appropriate for the H.E.S.S. site.Reuse & Permissions

Figure 4

Mean DC-ratio $Q_{\mathrm{DC}}$ as a function of the total intensity of the camera image $I_{\mathrm{tot}}$ for iron events from simulations (events which interacted in the atmosphere before passing the Cherenkov energy threshold were not considered since they contain no DC-light). The zenith angle is 0°. The error bars show the RMS of the distribution in each bin. The dotted line shows the selection function given in Table .Reuse & Permissions

Figure 5

A measured event with indications of DC-light in all four cameras images (indicated by arrows), after high threshold image cleaning. The reconstructed shower direction is shown by a cross ($\times$) in each image. The reconstructed energy of this event is $50/48\mathrm{TeV}$ based on QGSJET/SIBYLL simulations. The reconstructed impact parameter and DC-light intensity for each telescope are shown in the lower panels in each image. The energy and impact parameter resolutions are $\approx 20\%$ and $\approx 20\mathrm{m}$, respectively. The white points mark disabled pixels.Reuse & Permissions

Figure 6

Event rate for different telescope multiplicities $N_{\mathrm{tel}}$ compared to the expected rate from the simulation of the cosmic-ray flux using the SIBYLL and QGSJET interaction models. Additionally, the background event rate from the proton band, estimated using SIBYLL simulations, is shown. No background events were identified for $N_{\mathrm{tel}}\ge 2$ and upper limits are shown (this result is the same for both hadronic models, which is why only the SIBYLL simulations are shown). An upper limit at 95% confidence level is shown for $N_{\mathrm{tel}}=2$. The upper limits for $N_{\mathrm{tel}}>2$ are derived under the assumption that they have the same ratio to the detected event rate as the $N_{\mathrm{tel}}=2$ events. This is a conservative estimate since the fraction of misidentification is expected to decrease with increasing $N_{\mathrm{tel}}$.Reuse & Permissions

Figure 7

Mean reconstructed charge $Z^{*}$ as a function of the true charge $Z$ for DC-events in an energy range of $1.5<{log}_{10}(E/\mathrm{TeV})<1.7$ for both hadronic models. For clarity the x-axis is shifted by $+1$ for SIBYLL. The error bars show the RMS of the distribution in each bin.Reuse & Permissions

Figure 8

Comparison of shower parameter distributions for data and simulations of the cosmic-ray flux. The bigger panels show normalized distributions of $R_{\mathrm{core}}$, $I_{\mathrm{DC}}$, ${\Delta }_{\mathrm{DC}}^{\mathrm{dir}}$ and $X_{\max }$ for SIBYLL simulations and data. The smaller panels underneath the distributions show their mean values and the mean value for the different charge bands of which the simulated flux is composed. Additionally, they show the mean value for the distributions for the QGSJET simulations. When comparing the mean values for the simulated fluxes one should bear in mind that roughly equal contributions to the error bars come from statistical errors and from uncertainties in the reference composition, which is the same for both models. The systematic difference between the mean values for the shower maximum $X_{\max }$ is not unexpected, since this quantity is difficult to treat in the simulations (see text).Reuse & Permissions

Figure 9

$Z^{*}$ distributions of data (open circles) and the fitted SIBYLL model (black boxes) in five energy bands. The curves show the charge distributions of iron (dotted) and of the lighter charge bands (dashed) that compose the fitted model. The fit results are summarized in Table . Note that, as mentioned in Sec. 3d, the charge resolution improves with energy, observable by a narrowing of the $Z^{*}$ distribution of iron with increasing energy.Reuse & Permissions

Figure 10

Differential iron energy spectrum measured with H.E.S.S. for the hadronic models QGSJET and SIBYLL multiplied by ${\mathrm{E}}^{2.5}$ for better visibility of structures. The spectral points for both models are measured for the same energies. For better visibility the SIBYLL points were shifted 10% upwards in energy. The error bars show the statistical errors. The systematic flux error in each bin is 20%. The measurements from balloon experiments with data points at the highest energies are shown for comparison [8, 27, 28] (a compilation with more measurements from balloon experiments and space born measurements can be found in 1). For a better visibility no horizontal bars marking the bin ranges are shown, they can be found in the respective papers. When comparing the measurements one should bear in mind that the experiments have different charge thresholds for their definition of the iron band (see legend).Reuse & Permissions