How well do third-order aperture mass statistics separate E- and B-modes?

Abstract: With 3rd-order statistics of gravitational shear it will be possible to extract valuable cosmological information from ongoing and future weak lensing surveys which is not contained in standard 2nd-order statistics, due to the non-Gaussianity of the shear field. Aperture mass statistics are an appropriate choice for 3rd-order statistics due to their simple form and their ability to separate E- and B-modes of the shear. However, it has been demonstrated that 2nd-order aperture mass statistics suffer from E-/B-mode mixing because it is impossible to reliably estimate the shapes of close pairs of galaxies. This finding has triggered developments of several new 2nd-order statistical measures for cosmic shear. Whether the same developments are needed for 3rd-order shear statistics is largely determined by how severe this E-/B-mixing is for 3rd-order statistics. We test 3rd-order aperture mass statistics against E-/B-mode mixing, and find that the level of contamination is well-described by a function of $\theta/\theta_{\rm min}$, with $\theta_{\rm min}$ being the cut-off scale. At angular scales of $\theta > 10 \;\theta_{\rm min}$, the decrease in the E-mode signal due to E-/B-mode mixing is smaller than 1 percent, and the leakage into B-modes is even less. For typical small-scale cut-offs this E-/B-mixing is negligible on scales larger than a few arcminutes. Therefore, 3rd-order aperture mass statistics can safely be used to separate E- and B-modes and infer cosmological information, for ground-based surveys as well as forthcoming space-based surveys such as Euclid.