Origin of poorly conditioned susceptibility matrix within inverse CVM (Cluster Variation Method) algorithm in the case of Pd3V: A comparison to the case of Ni3V.

摘要

Transition metal alloys are interesting in order to test the Cluster Variation Method as a proper statistical model to describe order-disorder phenomena. In the case of such alloys electronic structure calculations all agree to predict short-ranged pairwise interactions as sufficient to account for phase diagrams and order-disorder properties in general. We have already reported (1-2) diffuse scattering experiments performed on the disordered state of two of these alloys Pd(sub 3)V an Ni(sub 3)V. Performing Inverse CVM within the 13--14 points approximation, we have determined pair interactions up to the fourth neighbors. In the process of determining these interactions, we have met some difficulties, especially in the case of Pd(sub 3)V where the experimental errors on the short-range order parameters were dramatically enhanced through the Inverse CVM process, leading to poorly determined interactions. In particular to precise these interactions, we had to use as supplementary informations i.e. the order-disorder transition temperature. In this paper, we would like to emphasize that these problems are closely related to the nature of the system, namely to its high degeneracy (1), and not at all, for example, to the quality of the experimental data. Moreover, we would like to show that the parameter responsible for this high degeneracy, namely the (100) antiphase energy is determined with a good accuracy compared to the individual interactions. We compare the case of Pd(sub 3)V to the Ni(sub 3)V one, for which the degeneracy is lifted. 7 refs., 1 fig., 1 tab. (ERA citation 15:034898)