Long-term relaxation of large–scale inhomogeneities in aluminum – rare earth metal melts with a content of the latter in the range of 5–10 at % was previously observed when measuring viscosity and density, but did not manifest itself when measuring electrical resistance and magnetic susceptibility. This behavior could be related both to the specifics of the measured properties and to the size of the samples, which in the case of viscosity and density is much larger: 12–15 g with a diameter of a cylindrical crucible ~15 mm for viscosity and density, and less than 1 g with a crucible diameter of 6 and 4 mm for electrical resistance and magnetic susceptibility, respectively. To solve this problem, the time dependences of the electrical resistance of the Al91La9 melt were measured at 1060°C for samples of various sizes by the rotating magnetic field method, namely for standard (crucible diameter of 6 mm) and enlarged (crucible diameter of 10.5 mm) When the sample increases, the random measurement error increases, so additional measures had to be taken to stabilize the current in the coils, creating this field. It was found that with an increase in the mass of the sample to 2.15 g with a crucible diameter of 10.5 mm the large-scale heterogeneity that occurs during the melting process is manifested. It is interpreted as a compact “cloud” of intermetallic microparticles surrounded by a melt with a high content of REM. The sample relaxes to an equilibrium homogeneous state in a few hours. To speed up the process, additional exposure is required – heating to a high temperature, about 1500°C, which reduces the time to less than one minute. At the same time, in small (0.7 g with a crucible diameter of 6 mm) samples, the mentioned heterogeneity does not occur. The probable cause of the inhomogeneity is the flow of REM atoms to the surface and the reverse flow of aluminum atoms into the volume during crystallization, which is similar to liquation during the crystallization of cast iron and steel. The measurements carried out allow us to estimate the scale of the resulting inhomogeneities, which corresponds to the size of the enlarged sample.
