Phase-Change Memories for Analogue In-Memory Computing
The research is motivated by the growing need to improve computational efficiency in Artificial Intelligence systems, where performance is increasingly limited by the cost of moving data between memory and processing units. A promising strategy to overcome this bottleneck is to perform computations directly within memory arrays, thereby reducing latency and energy consumption. Phase-Change Memories (PCM) represent a mature and advanced technology capable of high-density information storage and well suited for embedded and on-chip integration. This project aims to investigate how PCM devices can be exploited for analogue in-memory computing, linking their physical behaviour to computational performance. ; ; The main objective of the research is to study and explore PCM devices in depth, with particular attention to their suitability for analogue computation. The work will focus on understanding the behaviour of PCM cells, examining the correlation between the physical and chemical evolution of the phase-change material and the resulting electrical response. The project will also include the development of simulations to evaluate PCM performance and reliability under conditions relevant to in-memory computing. Neural network simulations will be carried out to establish ideal computational baselines and to compare them with the behaviour of real PCM arrays. In addition, the research will involve the electrical characterization of alternative memory technologies, such as EEPROM, in order to provide a fair and comprehensive comparison of their potential and limitations. The expected original contribution of the project includes new insights into PCM device physics, improved modelling and simulation frameworks, and a clearer understanding of the role of PCM in analogue computing architectures.
Back to Current Students