Electronic properties of GeSi/Si‐based metal oxide semiconductor field effect transistors (MOSFETs) are strongly influenced by the geometry, composition and strain state of the Ge containing source/drain stressor regions and the Gate channel in between. In the first part of our contribution, we hence present quantitative STEM Z‐contrast evaluations of the Ge composition in MOSFETs with different stressor geometries and verify the results with energy‐dispersive analyses of X‐rays (EDX) using the chemiSTEM system mounted on an FEI Titan facility. To evaluate the strain distribution within the gate channel and the stressors, we used nano‐beam electron diffraction employing a delay‐line detector (DLD, [1]) which enhances the acquisition speed for the 4‐dimensional data set (diffraction patterns as a function of the STEM raster position) to 10ms/image.
As illustrated in Fig.1 the top of the DLD consists of a microchannel plate (MCP) stack that causes a cascade of secondary electrons for each 300keV electron impinging on the detector. The heart of the DLD are 2 meandering wires shown in blue and red in which each cascade causes electrical pulses travelling towards the ends of the wires. Depending on the incident position of the electron a characteristic time delay between the arrivals of the 2 conjunct pulses at the ends of a wire is measured with high accuracy, giving the coordinate of incidence perpendicular to the meander. By crossing two such delay‐lines the point (x,y) and time τ of incidence can be detected. Thus the DLD allows for both the recording of a continuous stream of single electron events processed by a time‐to‐digital converter (TDC) with a time precision in the picosecond range and the in‐situ integration of the signal over a certain frame time to obtain conventional images. Note that no pixel raster is involved here. The two modes of operation are illustrated in the right part of Fig. 1.
An example for a MOSFET exhibiting 2 Ge regimes of 22% and 37% within the stressors is shown in Fig. 2 for the strain measurement along the [001] direction and in Fig. 3 for a measurement along the [110] direction. Here a STEM raster of 100x100 pixels was used with a dwell time of 40ms. Figure 2 was obtained by evaluating the position of the 004 CBED disc using the radial gradient maximisation method [2] whereas the 220 reflection was measured to obtain Fig.3. Two strain regimes are observed inside the GeSi stressors due to the different regimes of Ge contents. The gate channel exhibits compressive strain of up to 3% laterally and an expansion below 1% along [001]. Concerning the influence of the acquisition speed on strain precision, recording 10,000 diffraction patterns took 6.5min here, yielding a strain precision of 0.12% in terms of the standard deviation determined in unstrained Si regions. The same experiment at twice the scan speed (dwell time of 20ms) yields the same average strain distributions, however, at the precision of 0.18%.
In the second part, we report a systematic study on the epitaxy of GeSi on Si (111) as a function of growth temperature. In particular, temperatures of 400, 450 and 550°C have been used and the strains along [111] and [1‐11] have been measured. Using elasticity theory, composition maps were derived from the strain results, yielding the three Ge profiles in Fig.4. Obviously growth at 400 and 450°C leads to islands of pure Ge whereas the Ge content drops to 60% if the temperature is raised to 550°C. We finally discuss our results with respect to Si/Ge interdiffusion which causes a gradual change of composition at the interface.