Chemical and structural arrangement of the trigonal phase in GeSbTe thin films

Abstract

The thermal and electrical properties of phase change materials, mainly GeSbTe alloys, in the crystalline state strongly depend on their phase and on the associated degree of order. The switching of Ge atoms in superlattice structures with trigonal phase has been recently proposed to develop memories with reduced switching energy, in which two differently ordered crystalline phases are the logic states. A detailed knowledge of the stacking plane sequence, of the local composition and of the vacancy distribution is therefore crucial in order to understand the underlying mechanism of phase transformations in the crystalline state and to evaluate the retention properties. This information is provided, as reported in this paper, by scanning transmission electron microscopy analysis of polycrystalline and epitaxial Ge2Sb2Te5 thin samples, using the Z-contrast high-angle annular dark field method. Electron diffraction clearly confirms the presence of compositional mixing with stacking blocks of 11, 9 or 7 planes corresponding to Ge3Sb2Te6, Ge2Sb2Te5, and GeSb2Te4, alloys respectively in the same trigonal phase. By increasing the degree of order (according to the annealing temperature, the growth condition, etc) the spread in the statistical distribution of the blocks reduces and the distribution of the atoms in the cation planes also changes from a homogenous Ge/Sb mixing towards a Sb-enrichment in the planes closest to the van der Waals gaps. Therefore we show that the trigonal phase of Ge2Sb2Te5, the most studied chalcogenide for phase-change memories, is actually obtained in different configurations depending on the distribution of the stacking blocks (7–9–11 planes) and on the atomic occupation (Ge/Sb) at the cation planes. These results give an insight in the factors determining the stability of the trigonal phase and suggest a dynamic path evolution that could have a key role in the switching mechanism of interfacial phase change memories and in their data retention.