NEuroMorPHic devices bASed on chalcogenIde heteroStructures (EMPHASIS)

Present-day computers based on the von Neumann architecture are becoming increasingly inadequate for data-centric applications, such as artificial intelligence and machine learning, due to the physical separation of the processing and storage units. Neuromorphic hardware is a revolutionary approach that takes inspiration from the brain for surmounting this drawback: its basic building blocks are “artificial neurons”, which can serve both as computing and storage units.

In this project, we aim at creating novel prototypes of neuromorphic devices based on chalcogenide heterostructures. The heterostructures will consist of alternating thin layers of phase-change materials (PCMs) and confinement materials (CMs), such as transition-metal dichalcogenides (TMDs). By application of proper electrical pulses, PCMs undergo fast and reversible transitions between multiple states showing resistivity contrast. However, the variability and the temporal drift in resistance of the PCM cells hinder their use in neuromorphic devices. The presence of thin confinement layers has been shown to reduce both cell variability and drift, making these phase-change heterostructures (PCHs) excellent candidates for these applications.

So far, no systematic screening of CMs, PCMs, and their combinations has been carried out to optimize the properties of these PCHs. Besides the aforementioned cell variability, it is also crucial to improve their retention, as well as their energy consumption, switching behaviour and endurance. In this project, we will address these issues by combining computational screening with state-of-the-art experimental tools.

Starting date: March 23rd 2022 – Project duration: 3 years