(Invited) Challenges and Solutions in Metrology for Complex, Confined Systems

Abstract

Probing composition, doping , strain and defectivity in confined heterostructures (fins, nanowires, etc) represents a challenge in terms of 3D-spatial resolution, sensitivity as well as in localization and heterogeneity. We will discuss progress and limitations of two approaches. The first approach is linked to methods such as Atomprobe tomography or SPM imaging which provide ultra-high resolution inherently. Unfortunately when reducing the analysed volume to the atomic scale, signal intensity becomes very low limiting the achievable sensitivity. Moreover their application (for instance Atomprobe tomography on embedded FIN, STM on trenches embedded in STI) becomes very complex and prone to many distorting artifacts which one needs to overcome. Speed of the analysis and time to data becomes then another limitation. In the opposite approach one abandons the idea of achieving the resolution to probe inside one single device but rather exploits broad beams (>1-20 mm) to investigate an ensemble of many, small identical structures. In cases like self-focusing SIMS, the physics of cluster ion formation is exploited to confine the information to the very small regions of interest whereby the averaging over many features enhances the signal substantially and enables good sensitivity. Examples demonstrate composition analysis in trenches as narrow as 20 nm. The spatial selectivity even allows to determine the composition of two mixed phases (i.e. InGaAs and InAlAs). Similarly strain and composition can be measured with (Edge-enhanced) Raman in similar trenches, as the confined volume provides a 100x increase in sensitivity. Channeling RBS on ensemble structures has been exploited to probe the early stages of III-V in narrow trenches. The important relevance of this approach is that it provides the opportunity to study the material as it grows in such a confined volume which frequently is very different from blanket experiments while all conventional (broad) methods can still be used.