(Invited) ALD of Phase Change and Threshold Switching Materials for Next-Generation Nonvolatile Memory Devices

Abstract

Memory technology continues to be at a crossroads. Established devices such as DRAM, Flash, and SRAM remain dominant, but each are facing increasingly formidable challenges in scaling, cost, and performance. Emerging memory technologies are attempting to displace these three or to find application spaces between those they traditionally support. However, no emerging memory has been able to demonstrate significant advantages over the incumbents, resulting in a proliferation of options such as phase change memory (PCM), filamentary resistive memory (ReRAM), ferroelectric memory (FERAM), magnetic memory (e.g. STT-MRAM and SOT-MRAM), and several others. Of these, the most successful has been PCM as evidenced by the commercial introduction of PCM based crosspoint memory in 2017. PCM based crosspoint memory has had limited market penetration in part due to the high manufacturing cost; the current layer-by-layer approach to PCM fabrication makes scaling difficult. In order to be cost competitive, crosspoint arrays will need to be manufactured in a style similar to that of 3D V-NAND where the active layers are deposited conformally inside deep features simultaneously forming hundreds of junctions. Chemical Vapor Deposition or Atomic Layer Deposition of these films is likely required to make this a reality. Some progress in materials for the PCM memory element has been reported; however, a remaining concern for crosspoint memory is the selector device, used in suppressing the sneak path current, consisting of a non-linear two-terminal device to properly select each memory element. In addition to high non-linearity, a selector device also requires high on-state current density JON, fast switching time, high endurance, stability, high scalability and back-end-of-line (BEOL) integration. Several selector devices have been proposed to date, e.g., mixed-ionic-electronic-conduction (MIEC), field-assisted superlinear threshold (FAST), and amorphous silicon (a-Si) selectors, but Ovonic Threshold Switch (OTS) composed of chalcogenide multinary materials clearly appears as the best candidate to cover all the mentioned requirements for high-density crosspoint applications.Currently, chalcogenide materials are deposited by using physical vapor deposition (PVD) that allows for fine tuning of the film stoichiometry, but at the same time limits its conformality and homogeneity on a large scale. To address this challenge, atomic layer deposition (ALD) is desirable; unfortunately, only ALD OTS As-free compositions with high leakage and limited endurance have been developed so far. In this report we describe the process and materials for GeSbTe PCM and a novel quaternary GeAsSeTe ALD OTS composition. We show a physical characterization of the ALD films and provide a comprehensive electrical characterization of integrated devices in DC and pulsed regimes. The OTS selector performance, in particular, shows excellent results (See Figure 1) - good selectivity (> 104), low threshold voltage (Vth) drift (< 40 mV/dec), fast switching (< 10 ns), excellent endurance (> 109 cycles) and robust stability. These results pave the way for future development of ALD chalcogenide-based selectors as a leading technology for multiple-stack integration of crosspoint memory arrays. To provide a physical understanding of the devices under study, we propose an analytical model for the subthreshold conduction mechanism, allowing for future development and optimization of the composition for 3D crosspoint memory applications.Figure 1. Electrical performance of an OTS selector over one billion cycles a) Pulsed-IV taken at each decade of the 1e9 endurance cycles. b) Sub-threshold current measured at each decade of the 1e9 endurance cycles. c) Evolution of the threshold voltage throughout the endurance test. Each point is measured three times, and all three measurements are reported d) Selector window at each decade of the endurance test. Figure 1