The COMET-L3 experiment considers the long-term situation of corium/concrete interaction in an anticipated core melt accident of a light-water-reactor, after the metal melt is layered beneath the oxide melt. The experimental focus is on cavity formation in the basemat and the risk of long term basemat penetration. The experiment investigates the two-dimensional concrete erosion in a cylindrical crucible fabricated from siliceous concrete in the first phase of the test, and the influence of surface flooding in the second phase. Decay heating in the two-component metal and oxide melt is simulated by sustained induction heating of the metal phase that is overlaid by the oxide melt.The inner diameter of the concrete crucible was 60 cm, the initial mass of the melt was 425 kg steel and 211 kg oxide at 1665°C, resulting in a melt height of 450 mm. The net power to the metal melt was about 220 kW from 0 s to 1880 s, when the maximum erosion limit of the crucible was reached and heating was terminated.In the initial phase of the test (less than 100 s), the overheated, highly agitated metal melt causes intense interaction with the concrete, which leads to fast decrease of the initial melt overheat and reduction of the initially high concrete erosion rate. Thereafter, under quasistationary conditions until about 800 s, the erosion by the metal melt slows down to some 0.07 mm/s into the axial direction. Lateral erosion is a factor 3 smaller. Video observation of the melt surface shows an agitated melt with ongoing gas release from the decomposing concrete. Several periods of more intense gas release, gas driven splashing, and release of crusts from the concrete interface indicate the existence and iterative break-up of crusts that probably form at the steel/concrete interface.Surface flooding of the melt is initiated at 800 s by a shower from the crucible head with 0.375 litre water/s. Flooding does not lead to strong melt/water interactions, and no entrapment reactions or penetration of water into the melt did occur. A nearly closed surface crust forms within 60 s after start of flooding and quenches after not more than 140 s.Residual holes in the crust are closed and the crust separates the coolant water overlayer and the hot melt below. Concrete erosion continues -although reduced -with some 0.040 mm/s, and eventually the melt reaches the maximum erosion limit of the crucible, at what time the simulated decay power is cut off. Two volcanic melt eruptions, that developed at the crust surface for a duration of 2 min each, had minor influence on coolabilty, as only 2.8 kg oxide particles were ejected into the overlaying water pool. Finally, the volcanic vents were blocked, and there are no indications that water ingression occurred through the surface crust.ii Post test analysis of the solidified melt was performed after the crucible was sectioned. The solidified melt shows no indication of water ingression from the upper surface. Tight crusts explain poor heat removal to the flooding water and the ongoing concrete...