The core–mantle boundary is the only interface where the metallic core and the silicate mantle interact physically and chemically. Many geophysical anomalies such as low shear velocity and high electrical conductivity have been observed at the bottom of the mantle. Perturbations in the Earth's rotation rate at decadal time periods require the existence of a thin conductive layer with a conductance of 108 S. Substantial additions of molten iron from the outer core into the mantle may produce these geophysical anomalies. Although iron enrichment by penetration has only been observed in (Mg,Fe)O, the second dominant mineral in the lower mantle, the penetration process leading to iron enrichment in the silicate mantle has not been experimentally confirmed. In this study, high-pressure and high-temperature experiments were conducted to investigate the penetration of molten iron alloy into lower mantle phases; postspinel, polycrystalline bridgmanite and polycrystalline (Mg,Fe)O. At the interface between (Mg,Fe)O aggregate and molten iron alloy, liquid metal penetrated the (Mg,Fe)O aggregate along grain boundaries and formed a thin layer containing metal-rich blobs. In contrast, no penetration of molten iron alloy was observed at the interface between molten iron alloy and silicate phases. Penetration of liquid iron alloy into the (Mg,Fe)O aggregate is caused by the capillarity phenomenon or Mullins–Sekerka instability. Neither mechanism occurs at the boundary of pure polycrystalline MgO, indicating that the FeO in (Mg,Fe)O plays an essential role in this phenomenon. Infiltration of molten iron alloy along grain boundaries (capillarity phenomenon) is the dominant process and precedes penetration due to the Mullins–Sekerka instability. The capillarity phenomenon is governed by the balance of forces between surface tension and gravity. In the case where the ultralow velocity zone (ULVZ) with a low shear velocity is composed of Fe-rich (Mg,Fe)O, the maximum penetration distance of molten iron alloy by capillary rise is limited to 20m. The addition of iron-rich melt to the base of the mantle is therefore unlikely to be the main cause of the high conductance of the CMB region predicted from decadal variation of the length of day. Furthermore, the absence of molten iron alloy penetration into silicate phases does not allow an extensive modification of the chemical composition of the mantle by core–mantle interaction.