Analog-Based Synapse of Double HfZrO2 Ferroelectric FETs With Homogeneous Phase by Superlattice HfO2-ZrO2 Toward Energy Efficient Accelerator

Abstract

Since the analog-based energy-efficient accelerator for synapses is highly demanded in the artificial intelligence (AI) era, the homogeneous and coherence ferroelectric phase of HfZrO ₂ (HZO) by superlattice (SL) growth mode with double layers is proposed in this work. The experimental results demonstrate excellent linear alternating consecutive potentiation and depression conductance (α _p /α _d = −0.85/0.63) with V _RMS = 3 V. In addition, the proposed SL technique for HZOs validates the ferroelectric-based orthorhombic phase (o-phase) 75–79% by geometric phase analysis (GPA) compared to the solid-solution process for 62–64%. The double HZO (D-HZO) structure is employed for diverse coercive field (E _C ) distributions to exhibit multistate data storage with 8 identical gap V _T . The SL-DHZO has a sufficient ferroelectric domain, which is crucial to achieving the requirements of analog-based energy-efficient accelerators for synapses in computing in-memory generation.