An analytical expression for the concentration profile of a diffusing element partially soluble in the material's sample has been obtained on condition that the diffusion source is depleted with time. Examples of the use of the solution obtained for processing of diffusion experiments carried out with a number of impurities in beryllium have been considered. The use of the present model shows a more accurate agreement of the calculated and experimental concentration profiles, which enables one to refine the characteristics of diffusion mobility of the impurities in the materials under study.The stability of the structure of solid materials containing impurities is determined, as a rule, by the redistribution of the impurities between the solid solution and the isolated phases. The mobility of impurities in solid materials is limited by diffusion processes; therefore, the diffusion coefficients D i (i = 1, 2, ..., n) of the impurities, which are found by the corresponding experiments [1], are an important characteristic of any impurity-containing material. To increase the migration rate of the impurities one carries out the experiments at a higher-than-average temperature (homogenizes samples) and then extrapolates the result obtained for D to the region of low * ) temperatures, using the Arrhenius law [2]:Clearly, in extrapolating D(T) to lower temperatures, the error of determination of low-temperature diffusion coefficients increases; therefore, to improve the accuracy of their determination one must organize the processing of diffusion experiments so as to minimize the computational error for D.We recall that, in the course of diffusion experiments, one most often applies a source layer of labeled (radioactive) diffusing atoms to one side of the sample; the sample is annealed isothermally for a certain period; then one successively removes its layers on the source side of the source layer and analyzes the radioactivity of the sample's residue N(x), where x is the distance from the source layer to the sample [3].To simplify the processing of the experiment the layer applied to the sample is made as thin as possible and the geometry of the sample is selected so that the process of diffusion can be considered to be one-dimensional. The diffusion coefficient D is determined by comparison of the dependences N exp (x) obtained in the experiment and a certain reference calculated function N calc (x). The form of the latter depends on conditions that are realized at the boundary of the matrix and the source layer of a diffusing impurity in the process of diffusion. In [4], it has been shown that unreliable data on the boundary conditions reduce the accuracy of determination of the coefficient D several times. Consequently, extrapolating the result for D to low temperatures, one can make a mistake by an order of magnitude or more. In this connection, in processing the experiments, we seek to reconstruct the boundary conditions realized at the source layer-matrix boundary in homogenization of a sample as accurately as p...