Investigations on laser-induced phase transitions and multilevel switching in phase change materials and their suitability for phase change photonic memory applications

Abstract

The internet has revolutionized the digital world; as a result, the present-day computing systems are used extensively for searching and recording information. Therefore, a considerable amount of data is generated; the recent statistics show that 2.5 exabytes (1018) bytes are produced every day [1]. Further, it is estimated that it could reach 463 exabytes by 2025, where the significant contribution can be from the social media usage and other internet services [2]. In this aspect, high density memories are required to store this enormous amount of data. Hence, the recent trend of computing systems changes from computing centric to data-centric [3]. The working memory and storage memory are the two different semiconductor memory used in the von-Neumann based computing system. The frequent access of the program code and data required for the processor are directly fetched from the working memory, and hence it demands extended endurance cycles along with the fast operating speed. The reading and writing speed should be in the order of a few tens of nanoseconds, reducing the latency between the processor and memory. Based on the operating speed and storage density, there are two types of working memories: static random access memory (SRAM) and dynamic random access memory (DRAM). The SRAM has extremely fast operating speed, and hence it has been used as the cache memory in the central processing unit (CPU). However, the SRAM occupies larger footprints leading to low storage density. Therefore it is not a viable option to store a large amount of data. On the other hand, DRAM requires smaller area to store data and thus offers high data storage capacity. Nevertheless, DRAM consumes significant power upon refreshing at regular intervals because of its volatile nature. The hard disk drives (HDD) are considered as the storage memory, which is non-volatile and stores the data for a long time (~10 years) [4]. Furthermore, the storage memory shows high storage capacity as compared to the working memory. However, in recent years HDD has been replaced by NAND flash memories due to its improved operating speed. Nevertheless, NAND flash memory works at a low operating speed as compared to DRAM. Hence it cannot be used for working memory application. Besides, the other important issue related to NAND flash memory is the shortened endurance cycles. Therefore, there is a performance gap between the NAND flash memory and DRAM in terms of latency, endurance, and storage density [5]. As a result, there is an increased demand for the novel memory technology (figure 1.1), which could possess high scalability, non-volatility and reduced latency. The alternative non-volatile memory showing the combined features of both DRAM and NAND flash can serve as a storage-class memory (SCM), as shown in figure 1.1 [6]. These improved features of alternative memory technology can enhance the computation efficiency of von- Neumann computing systems.