Microstructure-dependent DC set switching behaviors of Ge–Sb–Te-based phase-change random access memory devices accessed by in situ TEM

Abstract

Phase-change random access memory (PCRAM) is one of the most promising nonvolatile memory devices. However, inability to secure consistent and reliable switching operations in nanometer-scale programing volumes limits its practical use for highdensity applications. Here, we report in situ transmission electron microscopy investigation of the DC set switching of Ge-Sb-Te (GST)-based vertical PCRAM cells. We demonstrate that the microstructure of GST, particularly the passive component surrounding the dome-shaped active switching volume, plays a critical role in determining the local temperature distribution and is therefore responsible for inconsistent cell-to-cell switching behaviors. As demonstrated by a PCRAM cell with a highly crystallized GST matrix, the excessive Joule heat can cause melting and evaporation of the switching volume, resulting in device failure. The failure occurred via two-step void formation due to accelerated phase separation in the molten GST by the polaritydependent atomic migration of constituent elements. The presented real-time observations contribute to the understanding of inconsistent switching and premature failure of GST-based PCRAM cells and can guide future design of reliable PCRAM. INTRODUCTIONPhase-change random access memory (PCRAM) devices store and erase information by utilizing a large resistivity difference between the crystalline and amorphous states of chalcogenide materials. 1 Among various chalcogenide materials, the Ge-Sb-Te (GST) alloy system is currently considered the most promising candidate material for PCRAM owing to its fast and reversible phase-transition capability, in both the amorphous-to-crystalline (set) and the reverse crystallineto-amorphous (reset) switching. 2,3 Moreover, the set switching of GST occurs through an abrupt increase in the current density at a critical electric field, which is known as threshold switching. 1,4 The threshold switching provides local Joule heat and thereby facilitates the crystallization process at a relatively low electric field, ranging from 30 to 50 V μm − 1 . 5,6 Many modern PCRAM cell designs adopt a nanometer-scale hemispherical or cylindrical shape for programing volume to increase the cell density and reduce the operation power. 7 In such device structures, only a portion of the programing volume makes direct contact with a heater electrode, and the other sides are surrounded by passive GST that remains in the crystalline state and acts as an electrical conductivity path during the set and reset switching. One of the critical reliability issues related to these cell structures is the compositional change, or phase separation, caused by electric field-

In situ TEM observation of void formation and migration in phase change memory devices with confined nanoscale Ge2Sb2Te5

“…According to eqn (1) , the amount of Joule heat generated within the confined cylindrical volume is determined predominantly by the σ and the k values for GST. We assumed that the GST changed to a crystalline state under the application of the set pulse of 2.5 V. All material parameters used in FEM simulations are identical to those used in the previous study, 29 and the physical parameters of voids are set to those of an air-like phase. The temperature map in Fig.…”

“…2b ). 29 The TiN BE was connected to an electrically grounded Cu support grid via the W plug. The set switching of the GST PCM device was driven by applying voltage pulses (2.5 V pulse for set switching and 4 V pulse for reset switching) during the observation of microstructural changes in real-time.…”

Direct observation of leakage currents in a metal–insulator–metal capacitor using in situ transmission electron microscopy

“…The MIM capacitor was composed of Ru/YSZ/Ru vertical stacked nanostructures. We have designed for the stacked thin film type to observe the Joule heating influence of the overall region under the top electrode rather than the electrical characterization in the contact area or the local electrode heater [25,26]. As shown in figure 1(b), the thicknesses of both Ru electrodes and the YSZ thin film are 10 nm and 20 nm, respectively.…”

“…In addition, we proceeded with the experiment for the biasing test when the probe contacted a thick Pt electrode that was not milled by the FIB. The re-deposition of Ga nanoparticles is also a critical problem as a pathway to conduct current during biasing tests because Ga ions are used in the FIB system [22]. To remove conductive Ga nanoparticles, additional milling with Ar ions was conducted using a NanoMill (Fischione, model 1040).…”

“…When a voltage is applied, Joule heating occurs due to the current flowing in conducting materials such as the Pt and Ru electrodes. The heating equation is [22]…”