CMOS-Compatible ScAlN Ferroelectric Thin Films with Enhanced Polarization for High-Performance FeFET Memory and...

Area of Science:

• Materials Science
• Solid State Physics
• Nanotechnology

Background:

• Scandium Aluminum Nitride (ScAlN) is a promising ferroelectric material for advanced electronics due to its high remnant polarization and CMOS compatibility.
• Conventional deposition methods face challenges with ScAlN films at high Sc concentrations and reduced thicknesses, leading to degraded ferroelectric performance and increased leakage.
• Developing stable, high-performance ultrathin ferroelectric films is crucial for next-generation low-power memory and neuromorphic devices.

Purpose of the Study:

• To investigate the impact of an optimized substrate structure on the ferroelectric properties of Physical Vapor Deposition (PVD)-grown ScAlN films.
• To demonstrate enhanced ferroelectricity in ultrathin ScAlN films with improved Sc concentration and crystal orientation control.
• To evaluate the performance of ScAlN-based ferroelectric field-effect transistors (FeFETs) and artificial synapses fabricated on the optimized substrate.

Main Methods:

• Physical Vapor Deposition (PVD) was used to grow ScAlN films on a novel optimized substrate structure.
• Ferroelectric properties, including remnant polarization (P_r), were characterized for Sc_0.2Al_0.8N, Sc_0.3Al_0.7N, and Sc_0.35Al_0.65N films at various thicknesses.
• ScAlN-based FeFETs and artificial synapses were fabricated and tested for memory window, switching ratio, retention, endurance, and neural network recognition accuracy.

Main Results:

• The optimized substrate significantly enhanced ferroelectric properties of PVD-grown Sc_0.2Al_0.8N films, retaining high P_r even at 20 nm thickness.
• Improved ferroelectricity was validated for higher Sc concentrations (Sc_0.3Al_0.7N and Sc_0.35Al_0.65N) across varying film thicknesses.
• The fabricated Sc_0.2Al_0.8N FeFET demonstrated a 17 V memory window, >10³ switching ratio, >10⁴ s retention, and >10⁴ cycle endurance.
• The artificial synapse achieved 98.7% recognition accuracy in neural network training, showcasing potential for energy-efficient computing.

Conclusions:

• An optimized substrate structure is critical for achieving superior ferroelectric performance in PVD-grown ScAlN films, especially at ultrathin scales.
• This advancement overcomes limitations of conventional methods, enabling stable single-phase ferroelectricity in ScAlN for demanding electronic applications.
• The demonstrated FeFET and artificial synapse performance highlights the potential of ScAlN for next-generation low-power, high-density memory and neuromorphic computing.