Disorder Control in Crystalline GeSb2Te4 Using High Pressure

Abstract

Electronic phase-change memory devices take advantage of the different resistivity of two states, amorphous and crystalline, and the swift transitions between them in active phase-change materials (PCMs). In addition to these two distinct phases, multiple resistive states can be obtained by tuning the atomic disorder in the crystalline phase with heat treatment, because the disorder can lead to the localization of the electronic states and, thus, hamper the electron transport. The goal of this work is to achieve and explore multiple disordered configurations in PCMs by applying high pressure. Large-scale ab initio molecular dynamics simulations demonstrate that pressure can lower the energy barrier for the antisite migration in crystalline PCMs. The accumulation of these antisite atoms largely increases the compositional disorder, adding localized electronic states near the conduction band. The disorder-induced electron localization triggered by pressure is a novel way to modulate the properties of materials. Furthermore, the random distortion of the lattice induced by the compositional disorder provides a new mechanism that contributes to the amorphization of crystalline PCMs at high pressure.

Keywords: Ge—Sb—Te (GST); ab initio molecular dynamics (AIMD); disorder; high pressure; phase‐change materials.

Figure 1

High‐pressure‐induced antisite hopping. We have performed AIMD simulations on c‐GST with 1008 atoms under various pressures. a,b) The atomic structures of c‐GST after 40 ps AIMD simulations without pressure and with moderate pressure (≈7 GPa). At moderate or high pressures, we observe a number of antisite jumps (the Te atom hops into an adjacent intrinsic vacancy site and the Sb atom then fills the empty site that the Te has left behind). The antisite Sb (Sb in Te layers) and Te (Te in Ge/Sb layers) are highlighted with green and pink spheres. The blue circles in the background mark the resulting ASPs, which result from the cooperative migration. Ge atoms are barely involved in such migration at moderate pressure. c) The percentage of ASPs (with respect to the total number of Sb atoms) increases with pressure. After 13 GPa, the crystal starts to turn into a glass.

Figure 2

Energy barriers for three migration pathways. NEB calculations provide the insight why antisite Sb/Te pairs can form under pressure, while neither single Te hops nor the formation of Ge/Te pairs are favored. The calculations were performed both for ambient and moderate pressure (≈7 GPa) at 0 K. a) The energy barrier that a single Te climbs over when it migrates into the neighboring vacancy. The pressure can reduce the energy cost of the migration but the energy increases monotonously, so that antisite Te does not find a metastable position at moderate pressure. b) The average energy barrier for the synchronized Sb/Te hopping. Te moves toward a neighboring vacancy, while Sb moves toward the empty site that this Te has left. Such cooperative migration has an average energy barrier of 1.1 eV at zero pressure and 0.9 eV at 7 GPa. The energy basin at the end of migration path explains why ASPs are metastable. c) The energy penalty of the synchronized Ge/Te hopping is rather high, indicating the low probability of the formation of antisite Ge/Te pairs.

Figure 3

Electron localization due to the antisite hopping. a) Davis–Mott model21 for typical density of states of disordered semiconductors. The disorder leads to the localization of electron wavefunctions near the extended (delocalized) states. In the c‐GST, the Fermi energy moves toward the valence band due to the defect formation on the cation sites. b,c) The density of states and IPRs of configurations without and with ASPs. The ASPs induce a peak in IPR near the conduction band (≈0.2 eV) in (c) suggesting that electron states (≈0.2 eV) are highly localized. Both extended states (without ASPs) and localized states (with 12 ASPs) near the conduction band are projected (in white isosurfaces) onto the real‐space atomic diagrams in (d) and (e), respectively. The pink, yellow, and blue spheres in (d) and (e) denote Ge, Sb, and Te atoms, and the red circles in (e) mark the positions of ASPs. IPRs and real‐space projections reveal that the ASPs indeed cause localized states near the bottom of the conduction band. The disorder of vacancies will also lead to electronic localization near the top of the valence band with and without ASPs.3

Figure 4

ASP‐induced random distortion in crystalline GST. To identify the distortion around ASPs, the partial PDF, which only considers the bond length around ASPs, is compared with the overall PDF. ASPs create some homopolar bonds such as Ge–Sb, Sb–Sb, and Te–Te, which have different bond lengths from heteropolar bonds. This leads to a broadened peak in the partial PDF, indicating that the bond length around the ASPs is less uniformly distributed.