Fermi surface incommensurate nestings and phase equilibria in Cu-Pd alloys Ezio Bruno, Beniamino Ginatempo, and E. Sandro Giuliano Dipartimento di Fisica and Unita ` INFM, Universita ` di Messina, Salita Sperone 31, 98166 Messina, Italy Received 20 December 2000; published 5 April 2001 The Cu-Pd phase diagram shows two low-temperature ordered phases corresponding to B 2 and L 1 2 super- lattices or, regardless of the site occupation, to bcc and fcc geometrical arrangements. We show how the phase equilibria between these intermetallic compounds can be understood by introducing a hypothetical bcc random-alloy phase and a three-step transformation from B 2 to random bcc to random fcc to L 1 2 . The two order-disorder transformations, B 2 to random bcc and random fcc to L 1 2 , analyzed in the framework of the concentration wave-functional theory, appear to be ruled by the sizes of the relevant Fermi surface FSnesting vectors in the two random-alloy systems. The transformation between the bcc and fcc solid solution phase is studied by first-principles Korringa-Kohn-Rostoker and coherent-potential approximation calculations, and it is shown to be influenced by electronic topological transitions occurring in the same systems. Our analysis points out the preeminent role played by the lattice-incommensurate FS nesting vectors that in both the random-alloy systems, give rise to frustrated concentration waves. The location of the B 2-L 1 2 coexistence at T =0 appears to be determined by the condition that the concentration waves of the fcc and bcc random alloys are equally frustrated. DOI: 10.1103/PhysRevB.63.174107 PACS numbers: 64.70.Kb, 71.18.+y, 71.20.Be I. INTRODUCTION In the last two decades, the first-principles determination of the phase diagrams of metallic alloys has stimulated a considerable amount of theoretical work. While remarkable progress has been made for order-disorder 1–4 and magnetic transitions, 5–7 the study of martensitic transitions remains more difficult. Although first-principles calculations are able to predict the most stable low-temperature structure, it is not easy to envisage from these sophisticated calculations which of the physical mechanisms are responsible for structural changes. For order-disorder and magnetic transitions, the Ginzburg-Landau theory allows us to determine, in the high- T homogeneous phase, the instabilities of the Fermi sea with respect to a certain infinitesimal charge or magnetization waves. Thus, perturbation theory can be applied and the seeds of the transition are identified by studying the response of the system to certain fluctuations. On the contrary, if the phase transitions have a peculiar first-order character, as in the case of most martensitic transitions, 8 perturbation theory is not applicable. We have discussed in a previous paper 9 a typical transi- tion for which perturbative approaches are expected to fail: the bainitic transition occurring in Cu c Pd 1 -c for concentra- tions 0.5c 0.7. In this case, the high- and low-T phases are, respectively, a fcc solid solution and a B 2 ordered su- perlattice based on a bcc geometrical lattice Fig. 1. The structural transition of the geometrical lattice can be thought as the consequence of a tetragonal deformation of the origi- nal fcc lattice. However, though one can imagine a continu- ous path, 10 the Bain path, along which the lattice is strained at constant volume with the tetragonal ratio c / a varying from 2 to 1 Fig. 1, nevertheless the strains involved are too large for applying perturbation theory, by including e.g., an elastic energy in the Ginzburg-Landau model. Thus, the only possible attack to the problem is the brute force total-energy calculation. This has been done in Ref. 9, where we have found a significant interplay between concentration fluctua- tions and lattice deformations. In short local chemical order- ing increases the charge transfer, which is quite large in or- dered Cu-Pd alloys, 11 hence, if a thermal fluctuation induces local ordering, then a considerable energetic gain can be ob- tained by local deformations that maximize the number of unlike neighbors. 9 As is well known, this is not possible in a fcc arrangement but it is in the B 2. The transition is, thus, at the same time, disorder-order and martensitic, and arises from the competition between the band energy and the elec- trostatics. For the present number of valence electrons per atom, 10.5e / a 10.7, in accord with the classical argu- ments of Mott and Jones, 12 the former favors the compact fcc phase, the latter the B 2 sometime called also CsClsuper- lattice, one of the typical structures of ionic crystals. A fur- ther reason for interest in this system is that another isoelec- tronic alloy Cu-Pt, at the equiatomic concentration, orders along a different direction, the 111, forming a L 1 1 superlattice. 13,14 We would like to add that the low-T phase diagram of a third isoelectronic system AgPd is not yet known and only a likely metastablefcc solid solution phase is reported in the literature. 13 The observed different phase behaviors of the above systems, all characterized by the same valence electron per atom ratio e / a =10.5, constitute an evi- dent violation to the Hume-Rothery rule 15,12 about the ‘‘band filling,’’ and suggest a scenario in which many different ef- fects compete with one another. The concentration range in which the Cu-Pd B 2 phase is observed at low temperatures is 0.5c 0.7, with a critical concentration c MAX 0.6. 13 On the other hand, recently pub- lished embedded-atom Monte Carlo simulations report a co- existence curve symmetric around c MAX =0.5. 16 As recog- nized by the same authors, their classical simulation fails to reproduce the experimental value because it cannot incorpo- rate the band-structure details. The wrong result c MAX =0.5 comes out from the inclusion of electrostatics and entropic effects only, since these are more important at the equi- PHYSICAL REVIEW B, VOLUME 63, 174107 0163-1829/2001/6317/1741078/$20.00 ©2001 The American Physical Society 63 174107-1