Soutenance de thèse : Camille LAGUNA

The Back-End Selector: from material development to device performances

Doctorante : Camille LAGUNA

Laboratoire INSA : INL UMR 5270
Ecole doctorale : ED160 Electronique, Electrotechnique, Automatique de Lyon

This work has its roots in the development of new memory concept in computer architecture with a goal to fill the gap between fast and enduring DRAM and non-volatile and dense Flash. With Storage Class Memories, we gather these performances. Thanks to the Crossbar architecture, 3D stackable in the back-end, we access a new range of portable memories. The memory revolution led by Intel open the door to 3D integration of memory on logic. This architecture is interesting for its low footprint (4F² per bit and per deck) which enable high density. Nevertheless, this requires a selector back-end integration. The back-end selector thus becomes a key device with reliability as main challenge. In particular, to open Crossbar technology to new applications such as neuromorphic computing for future embedded circuits. Devices reliability is linked to a good thermal stability compatible with BEOL integration, a very low threshold voltage variability to guarantee large reading window, an ultra-low leakage current to reach higher matrix sizes and an important endurance to reach applications compatible number of cycles. The thesis work is focused on OTS selectors (Ovonic Threshold Switching). They are based on chalcogenide glasses that present the following electrical characteristics: low leakage current in OFF state, high conductivity while in the ON state and volatile commutation between these ON/OFF states. A large range of materials based on SbSe and AsSe have been studied. The influence of additional elements such as germanium, silicon or nitrogen have been investigated. The material structure could be modified by alternatively depositing binary alloys creating multilayer materials. Their electrical and thermal performances are going beyond the ones of so-called bulk OTS alloys. A discussion is finally proposed and supported to increase the performances of future OTS materials.