INTERFACE SCIENCE 11, 33–40, 2003 c  2003 Kluwer Academic Publishers. Manufactured in The Netherlands. Effect of Grain Boundary Segregation and Migration on Diffusion Profiles: Analysis and Experiments J. BERNARDINI, CH. GIRARDEAUX AND A. ROLLAND Laboratoire Mat´ eriaux et Micro´ electronique de Provence, UMR CNRS 6137, Facult´ e des Sciences, St. J´ erˆ ome, 13397 Marseille Cedex 20, France D.L. BEKE Department of Solid State Physics, University of Debrecen, P.O. Box 2, 4010 Debrecen, Hungary Abstract. In many experimental studies, curved penetration profiles are observed for grain boundary diffusion performed in the B kinetics regime in contrast to the shape expected from the solutions of the second Fick’s equation. To explain these curvatures the effects of grain boundary structure, grain boundary migration, and grain boundary segregation have been successively proposed in the literature. Using previous data for Cu–Ag and Cu–Ni and new ones on Cu–Fe and Cu–Zn systems we will show how it is possible to separate all these possible contributions and how, knowing the true origin of the curvature, one can deduce much quantitative information impossible (or very difficult) to obtain by other techniques. Keywords: grain boundary diffusion, grain boundary segregation, radio-tracer 1. Introduction Grain boundary diffusion measurements are generally performed by using radiotracers coupled with a serial sectioning technique in the so-called B-kinetics regime [1] and the experimental data are interpreted assum- ing stationary grain boundaries, identical grain bound- ary diffusion coefficients in all grain boundaries and a Henry-type isotherm for segregating species [2]. Ac- cordingly, the logarithm of the specific activity versus the 6/5th power of depth is a linear function at dis- tances larger than 5 √ D v t [2], where D v and t are the volume diffusion coefficient and the annealing time, respectively. From the linear part of diffusion profiles the triple product P is determined, which is defined as: P = kD b δ, (1) where δ is the grain-boundary width (its value is gen- erally fixed at 5×10 -10 m), D b the grain-boundary dif- fusion coefficient and k the segregation factor. However, in many cases curved penetration profiles are obtained experimentally beyond the 5 √ D v t depth. Recent theoretical developments show that the inter- pretation of such complex penetration profiles should be based on more realistic assumptions than those listed above. In particular, one has to consider (i) the presence of different types of grain boundaries in the sample [3], (ii) the co-presence of moving and stationary bound- aries [4, 5] and (iii) non-linear solute segregation taking into account grain boundary saturation and interactions between segregating atoms [6–8]. As it is difficult to decide upon the dominant cause of the non-linearity of profiles from the shape analysis of experimental pro- files only, in the present paper we describe three sets of experiments using a radiotracer technique that make a distinction between the three different contributions listed above. For model systems, pure copper matrix with nickel, silver, zinc and iron as solute elements were chosen. Since nickel is completely soluble in copper [9] with a very small grain boundary segregation [10], effects other than that of the segregation can be studied, and