Emerging microelectronic applications such as Internet-Of-Things (IOT) or mixed signal processing utilize fully depleted MOSFETs built on SOI wafers. Advanced FD devices such as planar MOSFETs with back gate biasing require wafers with thin Si top layer as well as ultra thin BOX layer. To sustain good electrostatic control of the channel and to keep back-gate voltages reasonably low, top Si layer thickness has to be in the 5-10 nm range and BOX thickness needs to be less than 20 nm. The SmartCut^TM process is the only technology that allows transferring extremely thin layers of silicon on a wide range of handling substrates with industrial quality.

As threshold voltage of FD SOI transistor depends on the thickness of Si layer, the thickness has to be controlled within extremely tight tolerance, typically + /- 5A in full range of spatial frequencies spanning from 10 nm till 300 mm [1]. This creates challenge not only for manufacturing technology, but requires development of new thickness metrology. Introduction of differential reflection microscopy (DRM) [2] allows in-line SOI thickness control in wide range of spatial frequencies with unprecedented throughput and precision. The principles and main applications of DRM will be discussed in the presentation. The key process steps with their influence on specific spatial frequencies are identified. One of the main contributors to mid range frequency (in mm range) thickness variation is the interaction of propagating fracture front with acoustic waves emitted by itself. The experimental verification of this mechanism as well as the results of numerical simulations will be presented. Another critical process in SOI technology is reduction of high frequency thickness variations during finishing steps. Possibilities of using pure surface diffusion as well as chemically enhanced high temperature smoothing will be demonstrated. It will be shown that during high temperature annealing in neutral atmosphere Si surface evolution can be described by Mullings-Herring equation with two extra stochastic terms, responsible for conservative and non-conservative noise, while addition of HCl to the annealing atmosphere changes surface behavior to surface relaxation described by the Edwards-Wilkinson model. General linear 4^th order surface development model is proposed covering both mechanisms. Material parameters for the model are determined experimentally in the case of HCl(<1%)/H₂ etching of silicon.

For some advanced applications base wafer of SOI structure can play functional role in the final device, not limited to a mechanical support. One of such examples is RF SOI substrates specifically designed for RF analog applications, such as front end RF modules in smartphones. Introduction of thermally stable layer with high density of carrier traps between BOX layer and high resistivity Si substrate significantly improves high frequency electrical parameters of the devices built on such substrates [3]. Insertion loss, second harmonic generation of coplanar waveguides as well as quality factors of inductors are such main parameters. Empirical model is presented which allows prediction of the results of RF measurements from simple SRP data.

References

[1] T. Hook, et al, Electron Devices Meeting (IEDM), 2011 IEEE International (2011) p. 5.7.1.

[2] P. E. Acosta-Alba, et al, ECS Journal of Solid State Science and Technology 2, 357 (2013).

[3] J.-M. Le Mail, et al, Proc. VLSI-TSA 2015.