Machine learning method to predict threshold voltage distribution by read disturbance in 3D NAND Flash Memories

Abstract

Machine learning (ML) is proposed as a method to predict threshold voltage (Vt) distribution by read disturbance in the unselected strings of three-dimensional NAND Flash Memory (3D NAND). We extracted the Vt distribution after each read cycles in 3D NAND considering the process variation using Technology Computer-Aided Design (TCAD) simulation. The neural network (NN) was developed and was trained to have a small error rate. Through a test process, predicted Vt by ML was in good agreement with TCAD simulation data. In rapidly developed technology, the prediction by ML-based on the NN can be a powerful tool in terms of consuming less time. Also, ML can be applied to predict other conditions and reliability issues.