Testing a (quasi-)free base for modelling core-mantle boundary topography

The core-mantle boundary (CMB) is the most prominent compositional boundary inside the Earth. Its topography provides insight on lower mantle flow and the thermochemical structure above the CMB. Yet, CMB topography remains challenging to observe and estimates from seismology vary substantially. Numerical models of mantle convection provide complementary means to estimate CMB topography. Classically, topography is determined from the normal stresses acting on the CMB. However, this is known to face severe complications when applied to the surface boundary of the mantle, leading to non-Earth-like topographic scales and a different style of subduction. A (quasi-)free surface yields more Earth-like predictions, but for the CMB this comparison has never been made.

Here, we compare CMB topography predicted from mantle convection modelling using different treatments of the CMB. Specifically, we test the role of a ‘sticky core’, a quasi-fluid approximation the core. We compare results predicted by different codes (with either sticky core or true free base) and compare to a simple analytical case. Also, we simulate the evolution of subduction and deep thermochemical provinces to compare the topography of the (quasi-)free CMB and the free-slip approach. Initial results indicate that the sticky core approach can reproduce CMB topography reasonably well, but has rather high computational cost (grid resolution, number of particles). In analogy to the sticky air at the surface, the viscosity contrast of the sticky core layer determines the quality of predicted topography, with larger contrasts (≥10³) leading to acceptable levels of artificial CMB topography. In dynamic flow cases with vigorous mantle convection, entrainment by plumes further complicates application of the sticky core, but can be tackled with an unmixing procedure. A true free base tends to better accuracy than the sticky core approach and avoids the problem with entrainment, but it also comes with additional computational costs as various forces at the CMB have to be taken into account.