ScAlN is an emerging nitride ferroelectric material that exhibits exceptional remnant polarization (Pr) at ultrathin scales (<50 nm), stable single-phase ferroelectricity, and CMOS compatibility, making it highly promising for next-generation low-power, high-density memory and neuromorphic devices. However, ScAlN films deposited by conventional physical vapor deposition (PVD) faces challenges such as Sc precipitation and crystal orientation degradation at high Sc concentrations (>20%) and reduced thicknesses, leading to deteriorated ferroelectricity and increased leakage. In this work, it is demonstrated that an optimized substrate structure enables PVD-grown Sc0.2Al0.8N films to achieve significantly enhanced ferroelectric properties compared to conventional substrates, retaining high Pr even at 20 nm thickness. This improvement is further validated with Sc0.3Al0.7N and Sc0.35Al0.65N films across varying thicknesses. Additionally, a Sc0.2Al0.8N-based FeFET fabricated on this substrate exhibits a 17 V memory window, >103 switching ratio, >104 s retention, and >104 cycle endurance. When configured as an artificial synapse, the device achieves 98.7% recognition accuracy in neural network training under encoded pulse voltages, highlighting its potential for energy-efficient computing.
Keywords: ScAlN; artificila synapses; ferroelectric transistors; high‐performance ferroelectric film; neuromorphic computing.
© 2025 Wiley‐VCH GmbH.