Imprint issue during retention tests for HfO2-based FRAM: An industrial challenge?

Abstract

For ferroelectric random access memory (FRAM) with HfO2-based materials, the wake-up effect and the imprint have to be limited. Here, the electrical behavior of different samples is investigated during retention tests on woken-up samples at room temperature. Retention properties are compared during tests with or without alternations of voltage pulses with opposite signs. First, during retention tests with alternations, the imprint oscillates between two values that are believed to be too high for the reading operation of industrial FRAM memories. This imprint oscillation is not the sole cause of remanent polarization loss. Second, the wake-up effect and retention loss appear to be closely linked. Finally, two retention fitting models are tested: the first one follows a power law t−n and the second one corresponds to a stretched exponential behavior exp(−t/τβ). The data cannot be fitted by the power law at all, while the stretched exponential can fit the data after t>100 s. In fact, the stretched exponential model highlights that the remanent polarization reduction during retention tests can be separated into at least two parts: a behavior for a short period of elapsed time and a behavior after a long period of elapsed time. The origins of this two-part remanent polarization loss behavior are discussed.