Comprehensive Performance and Reliability Assessment of Se-based Selector-Only Memory

Abstract

Chalcogenide-based selector-only memory (SOM) concept has shown great potential for high-density high-performance applications. While very promising, the understanding of the physical mechanism behind the polarity effect is still lacking. This work provides an in-depth investigation of SiGeAsSe SOM device under various operating conditions in order to identify its potential limitations. Specifically, a trade-off between operating current (I<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">op</inf>) and programming speed is established. Additionally, cycling endurance is found to be limited by a destructive failure at high-I<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">op</inf>, whereas threshold voltage variability is a limiting factor at low-I<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">op</inf>. Moreover, the retention time and the temperature acceleration of memory window loss is studied, also revealing a pronounced I<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">op</inf> dependence. Finally, composition analysis identified no major atomic segregation associated with programming pulse polarity. Instead, an alternative model is proposed, capable of reproducing the polarity effect by treating OTS switching defects as anisotropic structural units.