Source number counts at high energies: Swift versus NuSTAR

The hard X-ray sky at energies above 10 keV has been extensively explored by the SwiftlGehrels and the NuSTAR missions in the 14-195 keV and the 3-24 keV bands. respectively. The main population of the hard X-ray detected sources are active galactic nuclei (AGN). A discrepancy has been reported between the number count distributions of the two missions in the sense that the extrapolation of the SwiftlGehrels number counts in the flux regime sampled by NuSTAR lies significantly below the NuSTAR counts. We explore this aspect anew by comparing the number count distributions derived from the SwiftlGehrels 105 -month catalogue with those from the serendipitous NuSTAR source catalogue. We use a novel number count distribution estimator that makes use of the C -statistic on the unbinned data. In the 14-195 keV band, the Swift/BAT counts follow a Euclidean slope with alpha = 1.51 +/- 0.10 (90% confidence level). The NuSTAR counts in the 8-24 keV band present a steeper slope with alpha = 1.7 1 +/- 0.20 because of an upturn at fluxes below similar to 2 x 10(-13) erg cm 2 S-1. The same upturn is observed in the soft (3-8 keV) NuSTAR number counts, which generally also present a steep slope with a = 1.82 0.15. Only the bright part of the NuSTAR 3-8 keV counts agrees with the Chandra number counts in the 2-10 keV band, while the fainter part (below similar to 7 x 10(-13) erg cm(-2) S-1) of the soft NuSTAR counts is in marked disagreement with the Chandra counts. Next, we compare the derived number counts in the different bands using our X-ray AGN population synthesis models. The comparison between the Chandra and the Swift/BAT number counts shows a very good agreement for the standard AGN spectrum with a power -law slope Gamma = 1.9, a high-energy cut-off at similar to 130 keV and a 2-10 keV reflection component of 3%. On the other hand, using this standard AGN spectral model, only the bright part of the NuSTAR 8-24 keV and 3-8 keV number counts agrees with the model predictions. It is therefore most likely that the disagreement between the X-ray number counts in the different bands is caused by the faint NuSTAR number counts. We discuss various possibilities for the origin of this disagreement.