A Novel Hybrid-FE-layer FeFET with Enhanced Linearity for On-chip Training of CIM Accelerator

In this work, a novel hybrid ferroelectric (FE) layer FeFET based analog weight cell was proposed to optimize the weight update linearity for on-chip training. Combining a Hf <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.5</inf> Zr <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.5</inf> O <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> and Hf <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.95</inf> Al <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.05</inf> O <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> layer, the hybrid FE stack features a more dispersed coercive field (E <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">c</inf> ) distribution, thus achieving a high linearity of αₚ=-0.48 and αd=-1.73, with a simulated ~ 92.4% recognition accuracy of the MNIST dataset. The mechanism of the improvement was analyzed by comparison with control group.