Fin-FET etching

300mm etching cluster (Figure 1)

The manufacturing of sub-10 nm FinFETs requires control of etching on a near-atomic scale that current reactors are not capable of achieving. To achieve this, it would be necessary to independently control the fluxes of radicals and ions and minimize the thickness of the reactive layers which drive the etching to < 1 nm (damage induced in the sublayer), which involves low energy ions. . In the JDP with AMAT, we characterized/developed new plasma sources at LTM and in the USA. The 300 mm cluster at the heart of the JDP is shown in Figure 1. Until 2015, we studied pulsed ICP plasmas and showed that it was possible to reduce gas fragmentation and lower ion energy by playing on the duty cycle. This made it possible to improve selectivity but only for a limited number of applications. Then in 2015, we developed (and patented) a breakthrough technology: “Smart Etch”, which does away with the concept of a reactive layer to improve the precision of the engraving without losing its anisotropy (Figure 2). The material is first implanted with light ions (H+, He+) which modifies its surface to a precise depth without etching it. Then this modified layer is etched by a remote NF3/NH3 plasma (without ions and therefore without damage). Installed at LTM in 2016, a prototype reactor makes it possible to carry out these 2 stages in the same chamber and thus cycle them.

"Smart-Etch" prototype reactor (Figure 2)

The studies on the mechanisms involved in the two stages mobilized the entire team. The capacitive reactor used for implantation was characterized in terms of ionic flux and ion energy, while the downstream plasma of NF3/NH3 was characterized by VUV absorption spectroscopy (detection of radicals including HF and NH4F involved in the formation of ammonium salts leading to etching). The mechanisms of selective etching were elucidated and it was shown that kinetic ellipsometry allowed excellent real-time control of the process. This approach is very efficient for etching the SiN spacers of FDSOI transistors and is promising for other applications. Finally, in 2017, the group's MD expertise made it possible to introduce another innovative technology (patented): the ultra-rapid modulation of reactive (Cl2) and non-reactive (Ar) gases. MD has shown that a reactive SiClx layer takes time (a few 100 ms) to reach a stable thickness: by stopping the plasma before, we can therefore control its thickness < 1 nm (which is done by cutting the injection of Cl2 gas) while using high energy ions and therefore without losing anisotropy. This technology has already been implemented on the latest industrial reactors, as discussed below as part of JDP research activities carried out directly at AMAT in the USA.