First-principles study on atomic and electronic structures of the phase change materials Ge2Sb2Te5

Abstract

Ge2Sb2Te5 (GST) has been widely used in phase-change random-access memory (PcRAM) owing to its high-speed and reliable phase transitions. However, the material property needs to be tailored further to overcome several technical problems. The low resistivity of crystalline GST requires high RESET currents to transform the material state from crystalline to amorphous. In addition, the thermal interference requires higher stability in the amorphous phase. As a way to solve this problem, doped GST materials have been actively investigated. However, the microscopic origin of doping effects has not been elucidated sufficiently. Furthermore, the atomic structures of amorphous GST have not been claified yet. For these purpose, we first study the atomic and electronic structures of the crystalline and amorphous GST using ab initio molecular-dynamics (AIMD) calculation. Then, the influence of dopants on the atomic and electromic structure of GST will be studied. For the last several years, the amorphous structure of GST has been studied intensively. It is intriguing that there is a critical discrepancy between experimental and theoretical observations. In experiment, the as-deposited amorphous structures are best characterized by the ideal glass, following 8-N rule. In theory, a series of molecular dynamics calculations, directly simulating the melt-quench process, have confirmed that the simulated amorphous structures deviate a lot from the ideal glass. To investigate the amorphous structure of GST that satisfies the 8-N rule, we performed alternative melt-quench simulations on Si2As2Se5 and replace atoms in the final structure with Ge–Sb–Te. The resulting structures have salient features of the 8-N rule such as the tetrahedral configuration for all Ge atoms and the localized Te lone pairs at the valence top. In addition, the average Ge–Te and Sb–Te distances are in good agreement with experiment. The energetic stability of the ideal glass supports the existence of this amorphous structure that is distinct from the melt-quenched glass. From the analysis of electronic structures and optical dielectric constants, it is concluded that the electronic character of the melt-quenched amorphous GST lies in between the resonant p-bonding of the crystalline phase and the covalent bonding of the ideal glass. Next, we studied the effect of dopants on the atomic and electronic structures of GST. The pronounced effects of dopants such as Si, C, N, and O atoms, on material properties of GST are investigated at the atomic level using ab initio calculations. In the crystalline phase, stable doping sites are determined by characteristic chemical bonds such as Ge–N and Ge–O. The comparison of lattice parameters between theory and experiment supports the existence of dopants at vacant or interstitial positions. The electronic density of states indicate that the localization at the valence top increases with doping, explaining the increase of resistivity in experiments. The amorphous structures of doped GST are obtained by melt-quench simulations and they are well understood by selective bonds between dopants and host atoms. The chemical bonds around dopants are more favorable in the amorphous phase than in the crystalline state, accounting for increased amorphous stability of doped GST. Among dopants that have been studied, C dopants are found to fundamentally alter the local order of amorphous network by increasing the population of tetrahedral Ge atoms significantly. In addition, the density of ABAB-type squared rings is much smaller than for the undoped case. The results indicate that carbon dopants are very effective in expanding covalent nature in amorphous GST and enhancing amorphous stability. The recrystallization of doped GST is also simulated and it is directly confirmed that the crystallization process of doped GST is slowed down by dopants.