Paper MS-ThM1
Measurement of GST Thin Film Composition and Thickness

Thursday, October 18, 2007, 8:00 am, Room 615

Recent studies demonstrated that phase change memory (PCM) can achieve fast switching speed, good reversibility and scalability, making it the most promising alternative non-volatile memory (NVM) technology for the next decade. Currently PCM is being actively studied in semiconductor industry, with Ge₂Sb₂Te₅ (GST) being the most widely investigated material. GST film thickness and composition are two important quantities that require close monitoring in manufacturing. We have experimented with a number of measurement techniques including x-ray reflectivity (XRR), ellipsometry, wavelength dispersive x-ray fluorescence (WDXRF), energy dispersive x-ray fluorescence (EDXRF), energy dispersive spectroscopy (EDS), Rutherford backscattering spectrometry (RBS), x-ray photoelectron spectroscopy (XPS), inductively coupled plasma optical emission spectroscopy (ICP-OES), and low energy x-ray emission spectroscopy (LEXES). The advantages and limitations of the various characterization and metrology techniques for GST film composition and thickness measurement are compared. XRR can measure GST film physical thickness and density independently with good precision. Our study of WDXRF shows that there is little interference between antimony and tellurium peaks, and that GST film thickness and composition can be measured simultaneously with good precision. XRR and WDXRF can often provide complementary information, which can be very helpful for troubleshooting process and hardware issues. More details will be presented in the paper. In EDXRF and EDS measurements, antimony and tellurium peaks overlap, and good peak modeling is required to deconvolute the peaks to ensure accurate measurement. RBS in general is a very good standardless analytical technique. However its use for GST application is limited since it can not resolve antimony and tellurium peaks. Particle-induced x-ray emission (PIXE) can be used together with RBS at the cost of measurement accuracy, yet PIXE still requires peak profile modeling. ICP-OES provides reasonable accurate results, but it is a destructive technique and thus not practical for process monitoring. We have also looked into XPS and LEXES for GST-related applications. The effects of sputtering process parameters such as pressure, temperature and wafer bias on GST film composition are discussed. The GST film composition variation through a sputtering target life is also monitored.