Non-Gaussian deflections in iterative optimal CMB lensing reconstruction

The gravitational lensing signal from the cosmic microwave background (CMB) is highly valuable to constrain the growth of the structures in the Universe in a clean and robust manner over a wide range of redshifts. One of the theoretical systematics for lensing reconstruction is the impact of the lensing field non-Gaussianities on its estimators. Nonlinear matter clustering and post-Born lensing corrections are known to bias standard quadratic estimators to some extent, most significantly so in temperature. In this work, we explore the impact of non-Gaussian deflections on maximum a posteriori lensing estimators, which, in contrast to quadratic estimators, are able to provide optimal measurements of the lensing field. We show that these naturally reduce the induced non-Gaussian bias and lead to unbiased cosmological constraints in $\mathrm{\ensuremath{\Lambda}}$ cold dark matter at CMB-S4 noise levels without the need for explicit modeling. We also test the impact of assuming a non-Gaussian prior for the reconstruction; this mitigates the effect further slightly, but generally has little impact on the quality of the reconstruction. This shows that higher-order statistics of the lensing deflections are not expected to present a major challenge for optimal CMB lensing reconstruction in the foreseeable future.