The results on the As-and P-diffusion in gallium melt are discussed. The definition method of P-and As-diffusion coefficients in liquid gallium was stated. This method was based on t h e analysis ofconcentration profile formed during the diffusion annealing with following tempering. The diffusion coefficients were measured on temperature region 11 13 to 1383 K. The expression for temperature dependence of diffusion coefficients was obtained.CHcTeMaX Ga-P II Ga-As. The diffusion mass-transport in the liquid phase is one of the most important factors determining the properties of A3B6 compounds and their solid solutions during the epitaxy process (SMALL and CROSSLEY).The information on diffusion coefficients in the melt is necessary for performing epitaxy under controlled conditions. The aim of the present paper is the description of the diffusion parameters measurements method and also the temperature dependence of diffusion coefficients in the systems Ga-P and Ga-As.There are two methods for determination of diffusion coefficients of phosphorus (DGa) and arsenic (DGa) in gallium melt. The first of them is based on indirect determination of the coefficients from the crystallization rate of semiconductors by KANENKO et al. and BRYSKIEWICZ. The second one is connected with the determination of semiconductors dissolution rate in the unsaturated melt by NEVSKY et al., and VIGDOROVrCH et al. The kinetic processes a t the interphase are not taken into account, whereas their contribution in the mass transport can be compared t o the mass transport in liquid phase. The convective mixing in the melt could also considerably influence the epitaxy crystallization rate (MILVIDSKI et al.).I n the present work the diffusion coefficients were determined directly from impurity (As or P) concentration profile formed after diffusion annealing and following tempering.The quartz capillar of 1.5e2.0 mm in diameter filled with gallium was used as an experimental cell. The impurity source in the form of single crystal (Gap or GaAs) was placed on gallium surface. The mass of the source was taken for the diffusion condition from the infinitely thin layer, Such niodel allows us to eliminate the influence of the processes at the interphase boundary. The vertical situation of the cell in the heater was used during the diffusion annealing. The impurity source was situated in the upper part of the capillar. This allows u s t o minimize the influence of convection mixing of the melt. The melt was undergone to the tempering at the liquid nitrogen temperature. After that the specimen was extracted from the capillar.

The investigation of diffusion processes in metallic melts is of great interest for liquid phase epitaxial development. However, there has not been until now reasonable concordance between experimental data on diffusion coefficients even for the most studied semiconductor systems. Evidently, this is connected with indirect approach to the solution of the diffusion problem. According to such an approach the calculation of diffusion coefficients is performed as a rule using the kinetic characteristics of solution or crystal growth of binary compounds in a melt. GOROKHOV et al. have pointed out three independent experimental possibilities for determining the impurity concentration profile : the electron probe microanalysis, XRSFA and phase transition temperature (PTT) technique. As has been stated all the results were in a good agreement with analytical dependencies of Gap-Ga and GaAs-Ga systems.I n the present paper the results on determination of As-and Sb diffusion coefficients temperature dependencies in In melt in the temperature ranges of 873-1227 K and 628-873 K are given. The diffusion coefficient has been found using concentration profile of As (or Sb) that has been formed during the diffusion annealing in indium melt followed by tempering a t liquid nitrogen temperature. The experimental cell was a quartz capillar of 0.8-2.0 mm in diameter and 00 mm long filled with indium melt. On the melt surface the impurity source InAs (or InSb) single crystal was placed. The mass of the source was taken for the diffusion condition from an infinitely thin layer into semiinfinite volume t o be correct. Such niodel allows us to eliminatc the limiting influence of interface phenomena of the impurity compound solution. After diffusion annealing the concentration profile was fixed up by tempering of the melt at liquid nitrogen temperature. The measurement of impurity content along the capillar axis was performed with XRSFA on fluorescent spectrometer VKA-2 (Carl Zeiss-Jena) using Cr anode. Before the measurement the original sample was cut in 8-13 probes. The length of each was 2-5 mm. The impurity concentration in each probe was estimated by a number of X-ray pulses during 20-100 s. ASK, and SbL, lilies were used for As-arid Sb analysis, respectively. BOURRET et al. have reported about unsuccessful attempts to use XRSFA for studying impurity diffusion in a capillar that was clue to small square of a sample uiider radiation. To solve the similar problem i n our caHe each probe was rolled for preparing a film sample with square under radiation of 1.5-2.0 om3. h e to high elasticity of indium this operation can be quite easily performed. The temperature dependeiicies of As-and Sb diffusion coefficients in indium melt obtained usiiig the technique described above are show11 in Figures 1 and 2. The impurity diffusion coefficients obtained using YTT-technique and others iiivestigation methods are also given in these figures.One can see that experiniental dependencies are possible t o be described by Arrhenius law that is: D = Do...