Thin Film Measurements with LEXS/HEXS

 

            WDS is particularly useful for making measurements on very thin films similar to what is used for semiconductor or optical applications because normally one knows the composition of the film and is trying to measure its thickness or stoichiometry.  The WDS can then be used to look at a specific element(s) unlike an EDS that absorbs data from everything including the substrate.  With WDS, you can crank the beam current as high as you like with no dead time issues.  LEXS/HEXS is more useful than a Rowland Circle WDS because it has much better performance for the films that are likely to be used such as nitrides and oxides.

            A good example is the measurement of very thin TiN films on silicon wafers as shown in Fig. 1

 

Fig. 1.  A very thin TiN film on a silicon wafer.

 

For this measurement, the old JEOL 733 Microprobe could not go to very low voltage but it does show that extremely good performance can be obtained at low voltage.  The N count rates from this thin film using this LEXS is better than a conventional WDS on a bulk sample. Note that the count rate drops with increasing voltage because most of the electrons are going right through the TiN film and into the substrate.  If we could have gone to sufficiently low voltage, we could have also looked at the Si substrate as a function of voltage and measured the thickness by comparison to standards or by using a program like Casino.

            Another good example of the utility of LEXS/HEXS for thin film measurements involved looking at thin TaN films under thick copper layers on silicon wafers.  In this case, any high enough energy electrons to go through the copper would go right through the thin TaN layer and into the Si.  Of course, there is an overlap between the Si(K) line and the Ta (Ma) line that would be resolvable if the TaN was not such a thin layer but even with LEXS, it is lost in the tail of the huge Si peak.  No other spectrometer would be able to do much with this situation but LEXS/HEXS is very good at seeing the N in the TaN, even through the thick copper as seen in Fig. 2.

 

Fig. 2.  N peak from 200 Angstroms of TaN beneath 1100 angstroms of copper atop a silicon wafer.