V-Ramp VBD Prediction Method Using OCD-Spectrum and Deep-Learning, and Application to Early Detection of V-NAND Low Metal Reliability Risk

For the first time, we propose a method to predict V-ramp V BD (Breakdown Voltage) using deep learning from OCD (Optical Critical Dimension)-spectrum. Using this, it was shown that the inter-metal dielectric V BD occurring in the LM (low metal) layers of V-NAND with the COP (cell-over-peri.) structure can be predicted. Modeling to predict V BD from OCD-Spectrum was performed using deep learning, and a highly accurate prediction model with consistency R2 of 0.78 and 0.58 was obtained in the modeling and mass data verification stages, respectively. Using this, V-ramp V BD can be predicted immediately at the process stage without waiting for fab-out, which is expected to contribute to improving mass production productivity by advancing defect detection.