On the influence of microstructure on fluorite ferroelectric embedded memories

The rise in artificial intelligence (AI) has led to a high demand for embedded non-volatile memories to reduce computation cost and power consumption. Fluorite-structure ferroelectrics have received growing attention with the discovery of ferroelectric hafnium oxide in 2008 and recent large-scale integration into industrial technology nodes, e.g. 1 Mbit ferroelectric RAM (FRAM) arrays by Sony. The ferroelectric layers are commonly of a polycrystalline nature, and due to the metastability of the ferroelectric orthorhombic phase, monoclinic phase fractions are often present. Since the crystallographic phase and its orientation will impact the observable polarization, the microstructure of the integrated film will be a major aspect for the device performance. Therefore, this work will explore the influence of the microstructure on different ferroelectric devices, i.e., ferroelectric field-effect transistors (FeFETs), ferroelectric metal field-effect transistors (FeMFETs), and FRAMs, and address its impact on vector-matrix multiplications used for hardware-accelerated AI.