The Earth’s lower mantle, ranging from 660 km to 2,890 km depth, constitutes more than 50% of Earth’s volume and is the largest geochemical reservoir for many elements. Throughout Earth’s history, substantial amounts of material have been exchanged between the deep mantle and Earth’s surface and atmosphere. Material transport from Earth’s surface into the deep mantle occurs by subduction of oceanic lithosphere. Historically, many scientists favoured a model of separate mantle convection within the upper and lower mantle, but seismology6, 7, mantle dynamic models8 and petrology9 provide evidence for subducting slabs penetrating the 660 km discontinuity and at least partial mixing between the upper and lower mantle. Recent seismic tomography studies suggest that in many subduction systems, such as under South America and Indonesia, slabs broaden and stagnate in the upper ~500–1,000 km of the lower mantle1, 2.
In high-strain regions of the lower mantle, such as near subducting slabs10 and in boundary layers11, (Mg, Fe)O ferropericlase—the rheologically weakest lower-mantle phase3—is likely to form an interconnected network12. In these conditions, the overall rheology will be dominated by ferropericlase even though (Mg, Fe)SiO3 bridgmanite (silicate–perovskite) is volumetrically more abundant in the lower mantle3, 12, 13. Previous high-pressure deformation experiments on MgO and (Mg0.83Fe0.17)O concluded that the rheological behaviour of (Mg, Fe)O is essentially unchanged throughout the mantle14, 15. In contrast to these experimental findings, an inversion of the energetically favoured slip system in MgO from 
110
{110} to 110{100} has been recently proposed to take place between 30 GPa and 60 GPa at 300 K on the basis of a multi-scale modelling approach16.
Here, we performed angle-dispersive high-pressure radial X-ray diffraction (rXRD) on powders of (Mg0.9Fe0.1)O and (Mg0.8Fe0.2)O at Beamline 12.2.2 of the Advanced Light Source, Lawrence Berkeley National Laboratory. Two runs were performed in the diamond-anvil cell (DAC) with a very fine pressure resolution to a maximum pressure of 83 GPa using X-ray transparent cubic boron nitride (cBN) gaskets. A third run was performed in a Be gasket to 96 GPa. All data were analysed for unit-cell parameters, elastic strains and texture (see Supplementary Information). A representative analysis is shown in Supplementary Fig. 1.
From our experiments, we derive differential elastic lattice strains. We observe an initial increase of all lattice strains Q(hkl), where hkl refer to the Miller indices of the respective lattice planes. This initial increase is followed by a saturation above roughly 10 GPa (Fig. 1), indicating that plastic flow was achieved. Starting from 20 to 30 GPa, our data show a second increase of all lattice strains up to pressures of ~60 GPa. Over this pressure range, we observe increases of 2–2.5 times for Q(111), Q(220) and Q(311), but only a slight change for Q(200). This pressure range is consistent with the pressure region of 30–60 GPa where changing slip system activities have been proposed for MgO by modelling16. In the following, we refer to the pressure range of increasing lattice strains as a ‘rheology transition’.