Former results in the molten-core-concrete interaction (MCCI) research indicate a possible penetration of a PWR cavity during a severe accident. To investigate the coolability of molten core the COMET facility was commissioned. The LACOMERA project is a 3 year shared-cost action within the Fifth Framework Programme. The overall objectives of the LACOMERA project are to provide research institutions from the EU member countries and associated states access to large scale experimental facilities which shall be used to increase the knowledge of the quenching of a degraded core and regaining melt coolability in the reactor pressure vessel, of possible melt dispersion to the cavity, of MCCI and of ex-vessel melt coolability. The experiment COMET-L2 is designed to answer following questions: investigation of long-term MCCI of metallic corium in cylindrical siliceous concrete cavity under dry conditions with decay heat simulation of intermediate power during test phase 1, and subsequently at reduced power during test phase 2. Observation of downward and sideward cavity erosion rates, cavity shape, and related processes. The experiment is performed in a cavity of siliceous concrete, initial inner diameter 595 mm. The mass of the metal melt is 430 kg, overlaid by 35 kg oxide melt. Decay heat was simulated by induction heating of the metal phase with a typical power of 200 kW, representing accident conditions in the late ex-vessel phase. Erosion of the concrete, temperatures of the melt, and erosion rates are measured. Unfortunately, the gas measurement system failed. In the first phase of the interaction, initial overheat of the melt is quickly reduced, while erosion rates into the axial and lateral directions are similar. When stationary conditions are achieved, the axial erosion seems more pronounced. Reasons for the differences between axial and radial erosion following the initial transient phase, and notably also the local inhomogeneities in axial ablation, may lie in the melt heating by a planar induction coil. This technique could possibly have resulted in inhomogeneous distribution of power in the melt, with a larger proportion close to the axial ablating concrete thus promoting axial ablation. The axial ablation might have been further amplified by possible positive feedback due to progression of the erosion front towards the coils, which eventually led to the cone-shaped erosion profile. Details of the experiment are reported to be used for validation of models and computer codes for safety assessment.