-
The air shower events are generated using CORSIKA v75000 [10]. Hadronic models EPOS-LHC [11] and FLUKA [12] are selected for interactions above and below 100 GeV, respectively. All shower particles have been tracked down to the threshold energy of 50 MeV for hadrons and muons and of 0.3 MeV for pions, photons, and electrons [13]. Five primary cosmic ray mass groups, i.e., H, He, C-N-O, Mg-Al-Si, and Fe, are simulated in the background estimation. The Poly-gonato model [14] is used for the fluxes and spectral indices of the species. The energy range of the incident cosmic rays is assumed to be from 10 GeV to 1 PeV for protons and helium and from 100 GeV to 1 PeV for the remaining three groups. Cosmic rays are assumed to isotropically arrive in the zenith angle range from 0
$ ^{\circ} $ to 60$ ^{\circ} $ . The energy spectrum of gamma rays is sampled from a power law SED with an index of 2.71 for the signals from the Crab Nebula. Every air shower event generated using CORSIKA is initiated at 100 random locations in an area of$2000 \;{\rm{m}} \times 2000\;{\rm{m}}$ centered at the center of WCDA-1 and followed by detailed simulation of the response of WCDA-1. More than 1.5$ \times10^{8} $ comic ray events and 2.7$ \times10^{8} $ gamma ray events have been generated in total. -
The code G4WCDA based on GEANT4 [15] has been developed to simulate the water Cherenkov detector response in the LHAASO experiment. To investigate the detector performance and correctly estimate the acceptance of WCDA-1, the detailed geometry and structure of the pond, including its roof, steel structure, beams, columns, concrete walls, detector suspension structure, and water transparency, are all taken into account in G4WCDA. Because approximately 300 Cherenkov photons are produced by an electron or a muon per centimeter of their tracks in water, most of the CPU time is consumed to simulate the huge number of Cherenkov photons for their propagation, including scattering and absorption in water. To improve the simulation efficiency, the propagation process is completed in two sequential steps, as sketched in Fig. 13. In the first step, photons and secondary particles in sub-showers are traced without taking into account the attenuation resulting from scattering or absorption. For only those that entered into a small volume of
$80\, {\rm cm} \times 80 \, {\rm cm} \times 80 \, {\rm cm}$ around the PMTs in each detector unit are all parameters of the photons and secondary particles collected and stored in a ROOT [16] file. In the second step, the attenuation is taken into account for the stored photons. The surviving photons and the secondary particles and newly generated photons are further traced until they reach the photocathodes of the PMTs. This approach has proven to be very effective in coping with various detector operation conditions, especially those associated with the water transmission, which was changing in the beginning phase of operation of WCDA-1.Figure 13. (color online) Sketch of the two step simulation procedure in G4WCDA. Left panel: Cherenkov photons/secondary particles were traced in water until entering the box in the vicinity of the PMTs. Right panel: Only photons or secondary particles that survived the attenuation cut were further traced until reaching the photocathodes of the PMTs.
The validity of the simulation was tested by comparing the distributions of several shower parameters with those measured in the observation. In Fig. 14(a) and (b), comparisons of the zenith (
$ \theta $ ) and azimuth ($ \varphi $ ) angular distributions between the simulated and measured showers with$ \theta<50^{\circ} $ are shown. Good agreement is evident, indicating that the attenuation of showers in the atmosphere as well as the uniform response of the detector array are well described. In addition to the variables related to the arrival direction, two shower energy-related variables are also compared, as shown in panels (c) and (d) of Fig. 14, i.e., the total charge Q and$N_{\rm hit}$ recorded by PMTs, respectively.Figure 14. (color online) Comparison between simulated (MC) and measured distributions: (a) distribution of zenith angle, (b) distribution of azimuth angle, (c) distribution of total charge in PE, and (d) distribution of number of hits
$N_{\rm hit}$ in the shower front. The vertical axes are event rates in arbitrary units. In d), simulation results using composition models proposed by Horandel [14] in blue and Gaisser et al. [17] in black are compared with the data.The shower core reconstruction described in Sec. III.A has been verified using the simulated events. The resolution is found to be 15 m, i.e., a circle with a radius of 15 m contains 68% of the content of the distribution of distances between the thrown and reconstructed core locations, for showers having more than 60 hits in the shower front.
To estimate the acceptance of WCDA-1 for gamma ray showers, the ratio of gamma ray showers surviving the C-cut for cosmic-ray background suppression has to be estimated correctly. The measured and simulated C distributions are compared in the left panel of Fig. 15. The agreement between the data and the simulation indicates that the algorithm is valid for the gamma-ray-induced showers from the direction of the Crab Nebula. In the right panel, the C distribution of selected events is compared with that for simulated showers with an assumption of a simple power law gamma ray spectrum of
$ E^{-2.71} $ . With a cut, e.g.,$ C>10 $ , a certain number of gamma ray showers are cut, while some cosmic-ray background events still survive after balancing between a signal-to noise ratio and sufficient gamma-like data statistics for later energy spectrum measurement. Simultaneously, the efficiency of gamma ray detection, or in other words the effective area, is also estimated in each bin of$ N_{\rm hit} $ , as listed in Table 1.Figure 15. (color online) Left panel: Distributions of compactness for data (in black) and simulated cosmic-ray background events (in red). Right panel: Distributions of compactness for selected gamma-like events and simulated events assuming a simple power-law gamma-ray spectrum
$ \propto E^{-2.71} $ .${ N}_{\rm hit}$ $E_{\rm med}$ $/{\rm TeV}$ Excess Background Significance (σ) Differential Flux/(cm−2 s−1 TeV−1) (a) 60 - 100 0.58 1438.2 24885.8 9.1 $(1.66\pm 0.20 ) \times 10^{-10}$ (b) 100 - 200 1.1 1082.7 5202.3 15.0 $(2.89\pm 0.23 ) \times 10^{-11}$ (c) 200 - 300 2.4 456.2 1376.8 12.3 $(4.74\pm 0.48 ) \times 10^{-12}$ (d) 300 - 400 3.9 161.2 335.8 8.8 $(1.12\pm 0.17 ) \times 10^{-12}$ (e) 400 - 500 5.9 60.3 77.7 6.8 $(3.54\pm 0.74 ) \times 10^{-13}$ (f) 500 - 800 12.1 82.7 45.3 12.3 $(6.91\pm 1.0 ) \times 10^{-14}$ Table 1. Summary of data used in the measurement of SED of the Crab Nebula over
$ 3.57\times10^{6} $ s. -
As mentioned above, the number of hits found within 30 ns of the reconstructed shower front,
$N_{\rm hit}$ , is selected as the shower energy estimator. For gamma-induced pure electro-magnetic showers, the simulated showers are used to establish the correlation between the primary energy and$N_{\rm hit}$ . In Fig. 16, primary energy distributions for gamma ray showers in the six bins of$N_{\rm hit}$ are plotted. Broad and quite overlapping distributions are observed. This indicates that the energy resolution of the detection with WCDA-1 is not perfect, in particular for showers whose major part may not be contained in WCDA-1. With WCDA-2 and WCDA-3 merged in, the detector sensitive area will be large enough to allow more strict event filtering, and the resolution could be improved for low energy showers in particular. The energy resolution improvement at low energy primarily comes from the reduction of edge effects. Here, the medians of the distributions are used as the measure of the gamma ray energy for showers in the corresponding bins of$N_{\rm hit}$ , as listed in Table 1. The energy resolution could be defined well as a symmetric Gaussian function above 6 TeV, where the resolution is approximately 33%. For lower energies, the resolution of${\rm log}_{10}E$ is better defined by a Gaussian distribution, e.g., with a width of 0.5 decades around 1 TeV. For showers having more than 800 detector units registered, the pile-up effect as shown in Fig. 9 makes the differential flux measurement impossible because of the natural saturation of$N_{\rm hit}$ at 900. An integrated flux above the energy corresponding to$N_{\rm hit} = 800$ could in principle be reported; however, the statistical significance is too weak,$ <2\sigma $ , to be reliable for the flux measurement here. Moreover, including a large number of units with too much charge recorded would require more careful tuning in the simulation to better simulate the response of the detector, as indicated in Fig. 14(d). The high energy events include both gamma-ray and cosmic ray induced showers, thus requiring a better energy estimator than$N_{\rm hit}$ . -
The SED of the Crab Nebula is measured using data collected from Sep. 5, 2019 to Feb. 29, 2020, for a live time of
$ 3.57\times10^{6} $ s. The criteria for events to be used in the analysis are as follows.● the core is within a square of 80 m
$ \times $ 80 m centered at the WCDA-1 center.●
$60 \leqslant N_{\rm hit} < 800$ .● The Crab Nebula was at least 45° above the horizon.
● The compactness is larger than 10.
All of these criteria helped to maximize the detection significance. The corresponding significance, number of events exceeding the background, and remaining cosmic rays in each bin are listed in Table 1. Under the cuts, WCDA-1 has an effective area that varies from approximately 400
${\rm m}^2$ at 600 GeV to a constant of 3200${\rm m}^2$ above 4.5 TeV, as plotted in Fig. 17. Importantly, this indicates that above 4.5 TeV, the SED measurement has a constant efficiency of approximately 50%, primarily becuase of the gamma selection requiring$ C>10 $ .Figure 17. (color online) Effective area of WCDA-1 for Crab detection with the cuts applied in the analysis according to the simulation for gamma rays. The red line indicates the physical area of the WCDA-1 used in the analysis.
The SED is determined by minimizing the
$ \chi^2 $ function$ \chi^{2} = \sum\limits_{i = 1}^6\frac{\left[N_{i}^{\rm obs} - N_{i}^{\rm exp}(\phi_{0},\alpha,\beta)\right]^2}{(\sigma_{i}^{\rm obs})^2} , $
(1) where
$N_{i}^{\rm obs}$ is the number of events exceeding the background in the i-th bin of$N_{\rm hit}$ ,$\sigma_{i}^{\rm obs}$ is the statistical error of$N_{i}^{\rm obs}$ , and$N_{i}^{\rm exp}(\phi_{0},\alpha, \beta)$ is the expected number of events according to the hypothesis of a log-parabolic model$ \phi(E) = \phi_{0}\left(\frac{E}{\;{\rm{TeV}}}\right)^{-\alpha-\beta\; \ln \left(\frac{E}{\;{\rm{TeV}}}\right)} $
(2) with three parameters (
$ \phi_{0} $ ,$ \alpha $ ,$ \beta $ ), where E is the shower energy,$ \phi $ is the photon flux from the Crab Nebula, and 3 TeV is taken as a reference point because WCDA-1 is the most sensitive at 3 TeV. This formula was suggested as one of the best functional approximations for the Crab nebula spectrum over a wide energy range [18]. The number of events for signals from the Crab Nebula direction is shown in Fig. 12 for each bin of$N_{\rm hit}$ with the background cosmic ray events subtracted. They can be fit quite well with a 2-dimensional Gaussian functional template.The SED measured using WCDA-1 is shown in Fig. 18 as the red points. The energy of each point is the gamma-ray median energy for the corresponding
$N_{\rm hit}$ interval. The corresponding differential flux in each bin is listed also in Table 1. The spectral parameters obtained in the log-parabolic fitting are$ \phi_{0} = (2.32\pm0.19)\times 10^{-12}{\rm{cm}}^{-2}\; $ $ {\rm{ TeV}}^{-1}\;{\rm s}^{-1} $ ,$ \alpha = 2.57\pm0.06 $ , and$ \beta = 0.02\pm0.05 $ with$ \chi^{2}/ndf $ = 0.513, respectively.Figure 18. (color online) The SED of the Crab Nebula in the energy range from 800 GeV to 13 TeV measured by WCDA-1 of the LHAASO is shown as the red filled triangles. As a comparison, the SED measured using KM2A of the LHAASO [19] above 10 TeV is shown as the green filled circles. The error bars are statistical errors, and the shaded area represents systematic error. Measurements of other experiments [20-24] are also plotted for comparison.
For comparison, the SED measured by other ground-based gamma ray experiments, i.e., HAWC [20], Tibet AS
$ _\gamma $ [21], and ARGO-YBJ [22], are shown in the same figure, together with the measurements by the IACT experiments, HESS [23], and MAGIC [24].The systematic errors are primarily caused by the following issues. First, WCDA-1 has been continuously improving in terms of the water transparency and detector stability throughout the observation period reported in this paper. This is reflected by Fig. 7(b), which shows that the measured average PMT charge generated by single muons nearly monotonically increased with time, except during major maintenance in October 2019. This requires a correction of approximately
$ \pm $ 11% to the charge measurements, thus introducing an uncertainty in the detection efficiency for events with a small number of hits, i.e., approximately 8.8% at$ Q = 0.3 $ PE, including non-uniformity among 900 detector units and variation over 170 operational days, tested using cosmic rays, which are uniformly distributed and hitting the detector from all directions. Second, the shower and detector simulation is an essential tool in estimation of the detector acceptance in terms of effective area for gamma ray detection. There is no way to check its validity directly using gamma rays because we do not have a known gamma ray beam for testing. However, it can be checked using cosmic rays. Assuming that the central region of WCDA-1, e.g., a circle with a radius of 30 m, is fully efficient for sufficiently large showers, e.g., showers with more than 500 hits, one can calibrate the absolute flux of cosmic rays calculated using the simulation codes in comparison with observations. Here, the details concerning the cosmic ray composition and the models of interactions with the atmospheric nuclei in the relevant energy range below 100 TeV have to be assumed as well. As shown in Fig. 14, there are discrepancies between the simulated and measured distributions. Here, two composition assumptions are compared with data in the$N_{\rm hit}$ distributions. This results in an uncertainty of the estimate of detector efficiency associated with the assumptions, including the entangled issues of interaction modeling and cosmic ray composition. The larger discrepancy of 16% between the two assumptions is taken as a conservative estimate of the corresponding uncertainty of the detector efficiency. The uncertainty of gamma ray detection acceptance estimated using the simulation tool is assumed to be the same in this analysis. Third, the gamma ray shower energy scale is also totally reliant on the simulation. As shown in Fig. 14(c), the simulation of the detector response near the threshold is slightly overestimated. To estimate the associated uncertainty of the energy scale, energies of the gamma events from the Crab Nebula direction have been reconstructed with thresholds of$Q_{\rm th}$ , 0.3 PE, and 0.6 PE. A difference of 3% is found between the average energies resulting from the two reconstructions; thus, a uncertainty of 8% is expected in the SED. The overall systematic uncertainty could be as large as$ ^{+8}_{-24} $ %. It is presented in Fig. 18 as the shaded area around the SED measured by WCDA-1.
Performance of LHAASO-WCDA and observation of the Crab Nebula as a standard candle
- Received Date: 2021-01-08
- Available Online: 2021-08-15
Abstract: The first Water Cherenkov detector of the LHAASO experiment (WCDA-1) has been operating since April 2019. The data for the first year have been analyzed to test its performance by observing the Crab Nebula as a standard candle. The WCDA-1 achieves a sensitivity of 65 mCU per year, with a statistical threshold of 5