Method of activation energy analysis and application to individual cells of 256Mb DRAM in 110 nm technology

摘要

In DRAM every memory cell experiences an individual mixture of leakage currents which consume part of the stored charge and lead to a wide distribution of data retention time (t_(Ret)). This distribution consists of an intrinsic (main) and an extrinsic (tail) branch. In this work, the method of activation energy analysis on individual cells is introduced for retention tail characterization and explained in detail. The formalism of activation energies (E_a) provides information about the mechanisms involved. Activation energies of single cells in a 256M DDR memory chip and their dependence on negative gate bias (VNWLL) as well as body bias (VBB) have been investigated intensively for the first time. Worst tail cells - all within a small retention time interval - show a twofold and wide distribution of activation energies. The lower E_a distribution can be altered with VNWLL, whereas the higher E_a distribution only alters with VBB. Going from tail towards main distribution, the percentage of cells belonging to the low E_a part continuously decreases and finally disappears. We therefore conclude that a gate induced mechanism (GIDL) is the main component responsible for DRAM retention tail.