Retention Improvement in Vertical NAND Flash Memory Using 1-bit Soft Erase Scheme and its Effects on Neural Networks

Abstract

We propose a selective 1-bit soft erase scheme in vertical NAND (V-NAND) flash memory that improves retention characteristics. Selective 1-bit erase using gate-induced drain leakage (GIDL) is applied after program operation to remove shallowly trapped electrons. Compared to conventional methods, the proposed method improves retention characteristics by ~40% and the distribution of $\Delta V_{\text {th}}$ is narrowed to less than 30%. The proposed $V_{\text {th}}$ tuning scheme accurately adjusts $V_{\text {th}}$ to the target $V_{\text {th}}$, thereby improving the error rate of convolutional neural networks (CNNs) for image classification by 17% by adopting the 1-bit soft erase scheme.