3D NOR-Type FeFETs with Record Endurance of 10¹¹, Fast Erase of 50 ns, and Immediate Read-After-Write for In-Memory Learning

In this work, we experimentally demonstrate a 3D NOR-type FeFET (FeNOR) array based on ALD IGO channel. With a scaled channel area down to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$0.007 \mu m^{2}$</tex>, the 3D FeNOR achieves a record endurance of 10<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">11</sup> cycles, 50 ns erase speed, immediate read-after-write, and minimal layer-to-layer variations. Comprehensive analysis unveils the mechanism of endurance degradation and optimization strategies. Based on multi-domain TCAD simulations, the improvement of erase speed is well-fitted with experiments and mainly attributed to blocking of percolation paths. The good density, reliability and speed make the 3D FeNOR promising for next-generation AI applications, especially in-memory learning.