Quantitative energy dispersive X ‐ray spectroscopy on thin S i G e layers

Abstract

The precise measurement of Ge content is of utmost importance in SiGe technology. Analytical methods like X‐ray diffraction (XRD) or secondary ion mass spectrometry (SIMS) allow the SiGe stoichiometry measurement in structures larger than 100 µm. However, for SiGe heterojunction bipolar transistors (HBT), Ge profiles in areas of typical transistor dimensions of about 100 nm are of interest. (Scanning) transmission electron microscopy ((S)TEM) in combination with energy dispersive X‐Ray spectroscopy (EDX) or electron energy loss spectroscopy (EELS) is one of the very few suitable methods for this purpose. Here, we present an approach for measuring Ge profiles in small areas with a lateral resolution of about 5 nm using the Cliff‐Lorimer method for quantification of EDX data of thin TEM samples and a calibration of the Cliff‐Lorimer factors to Ge profiles measured by XRD in large areas. We have investigated thin SiGe layers with thicknesses of about 20 nm and Ge content of about 30 at%. The Ge content was proofed by XRD as well. For EDX analysis, we have used a TEM FEI Tecnai Osiris operated at 200 kV in STEM mode. The EDX quantification was performed with Cliff‐Lorimer method using Esprit‐Software from Bruker. Measurements were taken on TEM sample with different thicknesses. The sample thicknesses were evaluated by EELS log‐ratio method in silicon area close to the SiGe‐layer [1]. Using this method, the specimen thickness is given in inelastic mean free path (mfp) units. The Cliff‐Lorimer method is widely used for quantification of EDX data of thin TEM samples. However, there is an uncertainty in the Cliff‐Lorimer factors for standard free measurements and the Cliff‐Lorimer method neglects any absorption effects which may become important for thicker TEM samples. Therefore it is necessary to proof the accuracy of quantification by using calibration samples with known Ge concentration. Figure 1 shows bright field STEM images of a Si 1‐x Ge x layer with x=0.305 (a) and a SiGe HBT (b). Figure 2 shows Ge line profiles of SiGe layer from figure 1a quantified using the K edges of Si and Ge measured at different TEM sample thicknesses. The obtained Ge concentration does not directly depend on sample thickness. However for very thick samples with a thickness of 2.73 mfp or 4.17 mfp, the apparent Ge concentration is clearly reduced. It is possible to explain the lowered concentration of very thick samples by stray radiation of surrounding Si and limited lateral resolution. Figure 3a shows the lateral resolution which was determined from line profiles using the method suggested by Williams and Carter in [2]. Points with concentration of 10 % and 90 % of the maximum concentration were measured and the distance between these points is multiplied by a factor of 1.8. The results from figure 3a and figure 2 clearly show that the sample thicknesses in the range 0.5 mfp to 1.0 mfp are a good compromise between lateral resolution and adequate signal‐to‐noise ratio. However, even for usual sample thicknesses of 0.45 mfp or 0.77 mfp, we have obtained a Ge concentration above 30.5 at% for the quantification with Ge K edge using a Cliff‐Lorimer factor of 2.386 and a slightly lower concentration for the quantification with Ge L edge. This suggests that an adjustment of Cliff‐Lorimer factors of Ge K edge and L edge for accurate quantified results of Ge concentration in SiGe samples is necessary. By using an adjusted Cliff‐Lorimer factor of 2.2 for Ge K edge and sample thickness below 1.0 mfp an error below ±10 % and a resolution of about 5 nm is achieved. Figure 3b shows a Ge line profile measured on SiGe HBT from figure 1b using EDX and quantified with adjusted Cliff‐Lorimer factor.