Achieving reproducible data: Examples from surface analysis in semiconductor technology

Repeatability and reproducibility in surface analysis in the semiconductor industry are key for supporting efficient process development and high volume manufacturing (HVM). Long-term repeatability is critically important when comparing to historical data, while reproducibility is required to support technology transfers when HVM of specific devices is to be carried out at multiple sites. This, however, introduces a number of unique challenges for running a characterization facility. In this work, the authors will describe several examples that can result in reproducibility issues. The examples will be taken in the areas of x-ray photoelectron spectroscopy and secondary ion mass spectrometry. The first and foremost origin of repeatability and reproducibility issues is instrumental variation. A second important contributor to reproducibility issues is sample variability. The authors will show that assessing long-term instrumental stability is potentially hindered by long-term variation of sample characteristics. The authors will also show that an understanding of characterization techniques is paramount to understand such issues. In addition to "pure" technical causes of repeatability and reproducibility issues, the human factor needs to be considered as well. This involves, for instance, decision making in data treatment during, for example, fitting procedures, statistical treatments, etc. Practical examples are given to illustrate this. With present day characterization depending more heavily on computational support/commercial software, potential detriments to characterization repeatability arising from software will again be made evident. Finally, the authors will illustrate with round-robin results that by combining all above-mentioned factors, widely varying results can be obtained on the same samples.