Scaling down the FD-SOI transistor size alone does not bring sufficient benefits to meet the performance roadmap. In the FAMES project, performance “boosters” are being researched and implemented to build FD-SOI next nodes.

One of the key boosters for 10 nm FD-consists in increasing the carrier mobility in the channel. The use of intrinsically strained channels is very effective to enhance carrier mobility.

Unfortunately, the strain required to increase nMOS and pMOS differ: tensile strained silicon is required for nMOS whereas compressive strained via SiGe channel for pMOS. The tensile layer (~+1.4 GPa) can be obtained at the wafer scale via sSOI strained SOI.

Such substrates show +33% Ieff gain as compared to conventional SOI substrates for nMOS devices. On the other hand, compressively strained-SiGe channel obtained by the Ge enrichment technique is interesting for pMOS devices.

However, in order to compensate or generate a 20% to 30% Ge condensed channel in sSOI wafer, a ratio of Ge around 50% epitaxy is required. A high Ge% could be difficult to integrate in devices (issues regarding threshold voltage adjustment).

To avoid this issue, it is preferable to relax the tensile stress of sSOI wafers before the condensation process. In the FAMES project, we have shown that by amorphizing the Si film on sSOI wafers by Si or Ar ion implantation, followed by annealing results in the relaxation of the tensile stress.

To enter further into the details, 3 implantation conditions (Si1, Si2 and Ar) have been tested. The goal of ion implantation is to amorphize a substantial portion of the thin sSi film while leaving enough crystalline regions on top to facilitate stress relaxation and subsequent recrystallization. All implanted wafers were then annealed at 850 °C for 5 minutes in a furnace to allow recrystallization down to the BOX.

This has been successfully demonstrated and confirmed by Transmission Electron Microscopy (TEM) and Raman measurements. The tensile stress can be reduced by up to 85%. Multiple iterations of implantation and annealing have been explored and can improve the stress relaxation.

Further studies will aim at improving the crystalline quality of the relaxed layers.