David Hellin

In metallic contamination analysis of semiconductor wafers, preconcentration methods such as vapor phase decomposition–droplet collection (VPD-DC) or droplet sandwich etch (DSE) are combined with sensitive analytical techniques, often total-reflection X-ray fluorescence spectrometry (TXRF). Previously we have reported about the occurrence of TXRF saturation effects in the analysis of microdroplet residues from these methods and identified the origin as a mass-absorption effect of primary X-rays. In this study, different approaches to deal with TXRF saturation effects in semiconductor applications are evaluated. Corrective approaches are tested for the application of VPD-DC-TXRF on Si wafers. The application of correction factors resulted in an overestimation of the effect of 20%. Internal standardization approaches did not perform well because of the separation of the internal standard element apart from the analytes during the drying step. We suggest another internal approach using the intensity of scattered primary X-rays as an indicator for accuracy.The parameters in a Ge matrix removal method have been investigated. The application of the method on samples from the DSE method on spin coated Ge wafers resulted in efficiencies at 100% compared to only 40% without matrix removal. Finally, the use of nanoliter deposition or ink-jet spray methods is proposed to extend the dynamic range of TXRF on microdroplet residues. This suggestion is motivated with calculations based on a mass-absorption model and with experimental results from the study of linearity of TXRF on Ni spin coated Si wafers and the related morphology of the surface species.

In future microelectronic devices, SiO₂ as a gate dielectric material will be replaced by materials with a higher dielectric constant. One such candidate material is HfO₂. Thin layers are typically deposited from ligand-containing precursors in chemical vapor deposition (CVD) processes. In the atomic layer deposition (ALD) of HfO₂, these precursors are often HfCl₄ and H₂O. Obviously, the material properties of the deposited films will be affected by residual ligands from the precursors. In this paper, we evaluate the use of grazing incidence- and total reflection-X-ray fluorescence spectrometry (GI-XRF and TXRF) for Cl trace analysis in nanometer-thin HfO₂ films deposited using ALD. First, the results from different X-ray analysis approaches for the determination of Hf coverage are compared with the results from Rutherford backscattering spectrometry (RBS). Next, we discuss the selection of an appropriate X-ray excitation source for the analysis of traces within the high- κlayers. Finally, we combine both in a study on the accuracy of Cl determinations in HfO₂ layers.

Total reflection X-ray fluorescence spectrometry (TXRF) is a well-accepted technique for ultra-trace analysis of ultra-pure reagents and silicon wafers. Nevertheless, the technique’s linear range is not well characterized. In this paper, the upper limits of the linear range of TXRF on micro-droplet residues are identified and the origin of the non-linear effect is investigated. It is observed experimentally that a systematic decrease in the accuracy occurs as a function of the metallic content, starting from amounts above 3–10 ng, depending on the sample composition. A mass-absorption model for thin films is re-formulated for the micro-droplet residue samples and the model parameters are tested on the experimental observations. The calculations are in good agreement with the experimental data. Finally, solutions to deal with these saturation effects are discussed and a method to extend the dynamic range of TXRF is proposed.