High speed nonvolatile memories employing ferroelectric technology

Abstract

The problems encountered in early attempts to build a ferroelectric memory are described. A combination of design innovations and process and materials breakthroughs that can overcome these problems is presented. PZT (lead zirconate titanate) is chosen as the basic ferroelectric material and integrated into a standard CMOS process. PZT has a wide temperature range (+350 degrees C Curie temperature), low coercive voltage, high specific polarization charge (10-20 mu C/cm/sup 2/), and good retention and endurance. The lack of a well-defined coercive field was overcome with a DRAM-like circuit architecture, which provides for transistor switches in series, with each ferroelectric element preventing disturb pulses from affecting the unselected cells. As a demonstration vehicle, a fully decoded 256-b nonvolatile ferroelectric random-access memory (FRAM) was developed. The switching speed inherent in the PZT material was found to be on the order of 1 ns. The high switching speed and the high signal charge, which render the technology highly scalable, offer the potential to build nonvolatile semiconductor memories with the speed of static RAMs and the density and cost of dynamic RAMs.<>