Spontaneous ordering in (Zn,Be)Se: A vibrational study by Raman scattering and first-principles calculations

The recent finding of a 1-bond→2-phonon behavior at the zone center in several A1−xBxC random zincblende mixed crystals, that mirrors a similar 1-bond→2-mode behavior in the bond length distribution, deviates strongly from the traditional 1-bond→1-mode (phonon, bond length) description of the local lattice relaxation in alloys, and thereby forces to rethink the scenario of spontaneous ordering (SO) in its entirety. Here we tackle this key issue, from both the experimental and theoretical sides, by focusing our attention upon the ‘model’ Zn1−xBexSe system. This alloy exhibits a large contrast in the bond properties, which should stimulate SO to some extent, and is at the origin of a uniquely well-resolved 1-bond→2-mode Raman behavior for the short Be-Se species, potentially a highly sensitive probe of SO. SO in Zn1−xBexSe (0<x≤0.92) is identified from the transverse optical Raman spectra as a slight but systematic reinforcement of the low-frequency Be-Se mode, that refers to the ‘longer’ bonds from the Be-rich region, to the detriment of the high-frequency Be-Se counterpart, due to the ‘shorter’ bonds from the Zn-rich region. This way the overall contrast between the Zn-Se (long) and Be-Se (short) bond lengths is reduced in the alloy, leading to higher crystal stability. The underlying mechanism of local Zn and Be segregation into a series of alternating cationic planes finds thus a natural explanation, which was still lacking. Direct insight upon the critical value of the order parameter η corresponding to ‘full condensation’ of the original 1-bond→2-mode (phonon, bond length) situation (η=0) onto the optimum 1-bond→1-mode end situation that minimizes the local strain energy in the representative ZnBeSe2 system (x=0.5) is achieved via first-principles bond length/phonon calculations in a series of fully relaxed Zn8Be8Se16 supercells with η ranging from 0 (random alloy) to 1 (CuPttype ordering). ‘Full condensation’ occurs at η∼ 0.5. Remarkably the local bonding between the anionic and cationic planes still has a (Zn-Se, Be-Se)-mixed character at this limit, as in a random alloy. In case of further ordering, the alloy would be forced to re-adopt a non-optimum 1-bond→2-mode type behavior, the result of the formation of ZnSe/BeSe micro-domains here and there. Basically η∼ 0.5 appears as an intrinsic limit to SO in ZnBeSe2, and possibly in stoichiometric alloys in general.