The use of design of experiments to evaluate the reliability of ferroelectric nonvolatile memories

A full factorial experiment was performed on 4 K ferroelectric random access memory (FRAM) devices to determine their fatigue performance as a function of fatigue temperature, fatigue voltage, read/write voltage, pattern, and number of cycles of fatigue. The resulting response surface of short-term retention (10 s at 80°C) and long-term retention (40 h at 100°C) and the product functionality were the metrics used to evaluate the product performance. An empirical model was used to predict yields and product performance. Based on this model, process modifications were undertaken to improve the performance of the product fabricated with PZT (nominal composition: Pb_1.10Zr_0.48Ti_0.52O₃. As a result, better than two-orders-of-magnitude improvement in product performance over operating conditions was seen. In addition, a statistically based monitoring program has been developed to ensure product performance