Electrothermal Modelling of Phase Change Memory Devices for Analog In-Memory Computing

Nowadays the digital era is becoming increasingly demanding on memory devices: higher memory capacities that are both reliable and cost-effective are needed for big data applications. Flash-based storage devices are now dominating the market, but they exhibit scalability limits. Researchers have started to explore new opportunities, looking at emerging memory devices, mainly non-volatile, that can replace or complement the actual storage technologies. The goal is to release the processor performance that is now constrained by the memory access and power consumption of the memory unit. Phase Change Memory (PCM) is considered one of the major candidates for such a purpose. PCM is, indeed, a non-volatile memory recently exploited for nanometric data storage and non-Von Neumann computing such as in-memory computing and neuromorphic computing. It consists of a layer of phase-change material sandwiched between two electrodes; the volume should be small enough to guarantee a low power switching capability. The phase of the material is switched by applying electrical current pulses that lead to an increase in temperature through Joule heating (write operation). Data is stored by exploiting the electrical resistance contrast between the low-conductive amorphous phase HRS (High Resistance State) and the high-conductive crystalline phase LRS (Low Resistance State) of the phase change material. Applying a very low power signal, the device's electrical resistance is measured, and one can easily retrieve the stored data (read operation). The main objective of this thesis is to develop an electrothermal model to simulate the behavior of PCM cells using COMSOL Multiphysics, a software that enables the modeling of coupled physical phenomena such as heat transfer and electrical conduction, both of which are critical in PCM devices. To achieve this, I developed a new simulation methodology to obtain the programming characteristics of the devices, linking COMSOL to MATLAB, which allows for quick comparison of different designs. The conventional ‘Mushroom’ configuration of the device is used as a starting point and reference for discussion. Specifically, the model is tailored to match the programming curves of the industrial PCM devices, fitting experimental data. The influence of geometric scaling and material parameter variability on the device’s behaviour are analyzed and reasoned by invoking various transport mechanisms, with the goal of improving performance. To achieve this, new device concepts are also considered and compared with the well-established mushroom configuration, such as the ‘Disc-Type’ cell, which is currently the focus of ongoing research.