TEM/STEM investigation of the crystallization kinetics of GeTe and Ag4In3Sb67Te26

Abstract

Phase‐change materials are promising candidates for non‐volatile electronic memory applications already successful established as rewritable optical data storage (Raoux, 2009; Meinders et al., 2006). The writing speed and the stability of the phase‐change data storage is mainly defined by the crystallization kinetics. Therefore knowledge of the crystallization kinetics of phase‐change materials is mandatory. Lee and coworkers have demonstrated by fluctuation electron microscopy (FEM) that the crystal nucleation of Ge 2 S 2 bTe 5 and AgInSbTe is influenced by their medium range order (MRO) of the amorphous phase (Lee et al., 2014). In our work we measured the crystal growth velocities by brightfield transmission electron microscopy (TEM) and we measured the variance of the diffracted intensity as measure of the MRO by FEM to prove a relation between the MRO and the growth velocity of the investigated phase‐change materials. The measurements of the growth velocities and the MRO were done for GeTe and Ag 4 In 3 Sb 67 Te 26 (AIST) in the amorphous sputtered as‐deposited and in the amorphous melt‐quenched state (Figure 1 and Figure 2). The 30 nm thick amorphous GeTe or AIST layer is embedded in a supporting layer stack on a 500 µm thick silicon substrate (Figure 2). The 100 nm thick ZnS‐SiO 2 capping layer on top of the AIST layer prevents oxidation. The 10 nm thick ZnS‐SiO 2 layer below the phase‐change material layer decreases together with the capping layer the necessary power to melt the phase‐change material layer by laser irradiation. The 50 nm thick Si 3 N 4 layer is an etch stop needed to prepare the TEM specimens by etching. The silicon substrate delivers sufficient heat dissipation for melt quenching. The GeTe or AIST layer is either investigated in the amorphous as‐deposited or in the amorphous melt‐quenched state. Melt‐quenched amorphous marks are produced by cooling a laser molten area of the phase‐change layer rapidly back to room temperature. The growth velocities were measured by brightfield TEM in a FEI Tecnai F20 at 200 kV. Electron transparent specimens were prepared by mechanical grinding, dimple grinding and etching with KOH from the layer stack described above. In this process parts of the Si substrates were removed to leave few hundred µm in diameter wide homogenously thick electron transparent windows. The specimens were alternated between heating in a heating furnace of a differential scanning calorimeter or heating in an oil bath for higher temperatures and the TEM to measure the size of the imaged grains. The measured grains were fitted by a circle. From a linear fit of the increase in radius of the fitted circles the growth velocities were calculated. FEM measurements were conducted in a FEI Titan dedicated to scanning transmission electron microscopy (STEM) at 300 kV (Heggen et al., 2016). A coherent almost parallel around 2 nm in diameter sized electron probe was used to generate the 500‐1000 nano area electron diffraction patterns per specimen used for the FEM analysis. The measured TEM specimens of as‐deposited and melt‐quenched GeTe and AIST were prepared from the layer stack described above as cross section lamellas by focused ion beam (FIB). The lamellas were produced in a FEI Helios dual beam scanning electron microscope/FIB system.