Reliable Local Strain Characterization in Si/SiGe Based Electronic Materials System

Abstract

In this research we first developed a procedure to determine the strain in a TEM sample. This procedure includes HOLZ line detection from a Convergent beam electron diffraction (CBED) pattern, kinematic calculation of high order Laue zone (HOLZ) line position and searching lattice parameters by χ2 minimization. With only CBED technique, strain measurement on the strained Si layer is not possible in a blanket strained Si⁄SiGe structure due to HOLZ line splitting (deformation). For sub-100nm short channel SiGe CMOS device structures strain could be determined in the center of the channel. We demonstrated the CBED strain measurement can be implemented in new generation short channel technology node with a nano meter spatial resolution and high accurate.

For the first time, we developed a new approach combined with CBED and finite element (FE) modeling and quantitatively investigated the correlation of the strain in a thin TEM sample with that in the bulk. The new method successfully determined the strain in the strained Si layer on a blanket strained Si/SiGe wafer, in a good agreement with other measurements. The new results also gave some insight in strain relaxation in a TEM sample. We found the [-1,-1,0] strain component which is perpendicular to the TEM sample thinning direction stays the same in the TEM sample and in the bulk, while the [001]) strain component is relaxed because it is along the same direction as the TEM sample thinning direction. This relaxation causes the deformation of the TEM foil and HOLZ line splitting. Therefore a clear CBED pattern can not be obtained from a TEM sample with a biaxial stain state. Our findings from a recessed SiGe PMOS test structure with a uniaxial compressive strain showed a different strain redistribution behavior. The data showed that the εx [-1,1,0] strain is actually more than 20% higher in a TEM sample than in the bulk. The εy [-1,-1,0] strain which is parallel to the TEM sample thinning direction turns to tensile in the TEM sample due to the loss of constraints, while it is zero in the bulk. The new results can explain our experimental data and others' (which could not be explained before) and are consistent with UV Raman measurements.