Resistive Switching Memory and Reconfigurable Devices.

Abstract

The demand in the data storage has been driving the explosive development of non-volatile memories (NVMs). Based on the field effect transistor (FET), flash memory has benefitted from the geometric scaling and dominated the electronics market. However, as Moore’s Law approaches its limit, flash memory is facing challenges from technical issues and economic concerns. Therefore, for a feasible next-generation NVM solution, a novel non-FET based replacement of flash memory is urgently desired. Among different candidates, resistive switching memory (RRAM) has attracted broad interest. In this work, we present studies on RRAMs and related reconfigurable devices. First, we systematically investigate the “sneak-path” issue of crossbar RRAM arrays and specify the selector device requirement in the one-selector-one-resistor (1S1R) configuration. We analyze the crossbar array from a perspective of device-circuit interaction and propose optimized benchmarks. Next, we develop a tantalum oxide (TaOx) based selector with high nonlinearity (~10000) and good uniformity. The conduction mechanism of this selector is found to be thermionic emission and tunneling emission. A HfO2 switching layer is integrated with the proposed selector to constitute a self-rectifying RRAM cell with high LRS selectivity (~5000), which can potentially enable Mbit crossbar array. Further, we demonstrate sub-nA operation current in a Cu based conductive bridge RAM (CBRAM) device for the first time, which offers significant energy savings during program and read steps. An improved cell with a built-in barrier/rectifying layer is developed to enhance device reliability. Apart from low current, other attractive properties including high on/off ratio (>100x), retention (over 10000 seconds at 100°C) and endurance (500 cycles without external current compliance) can be obtained. Additionally, we explore coupling the ionic migration process in resistive switching devices with transistor operation. A reconfigurable top-gate transistor is developed for the LaAlO3/SrTiO3 two-dimensional electron gas (2DEG) system. By incorporating ionic processes in the gate stack, we show that the channel conductivity can be modulated in a non-volatile manner by an external electric field. Finally, we propose a novel in-memory computing architecture using crossbar RRAM arrays, which breaks the boundary between computing and memory.