Total reflection X-ray fluorescence spectrometry (TXRF) has become one of the most common tools to analyze metal contaminants on silicon wafers and other substrate surfaces.1 The lower limit of detection (LLD) of transition metals by TXRF was 10 12 atoms cm -2 or higher at the end of the 1980s. It reached below 10 10 atoms cm -2 early in the 1990s. 2 Since then the TXRF came into wide use in semiconductor analysis.The accuracy is one of the most important factors for the reliable quantitative analysis of transition metals on Si wafer surfaces when we perform the TXRF analysis. In order to standardize the TXRF measuring method to ensure the accuracy of TXRF among different laboratories and/or instruments, the Ultra Clean Society (UCS) carried out some interlaboratory collaborative studies (so called "round robin") and reported a standardization document.3 In its round robins, the deviation of determination values of unknown samples was large when referring to individual standard samples prepared by each laboratory, and it became quite small when referring to fixed standard samples prepared with an adequate method by one manufacturer. From these results, the UCS reached the conclusion that the calibration standard sample was of critical importance for the accuracy of quantitative analysis by TXRF. The UCS round robins were, however, not entirely satisfactory because of the two points which did not reflect the present status of semiconductor analysis. The first fault was that the concentration tested was too high to apply the UCS results to an actual sample. Because the UCS focused on gathering the basic information, the concentration of the distributed samples was mainly above 10 12 atoms cm -2 to eliminate the statistical error of individual TXRF measurements. Accordingly, the lowest concentration of samples was 8´10 10 atoms cm -2 , in spite of the fact that the surface metal contamination of cleaned wafers in recent years never exceeds 5´10 10 atoms cm -2 because of the technology improvements of wafer cleaning and chemistry purification. The accuracy of quantitative analysis using TXRF at such lower concentration has not yet been clarified by any organizations or researchers. The second fault was the insufficient choice of the analyte element. The UCS round robin was carried out mainly for Ni, because Ni on silicon wafer was stable and was rarely contaminated from the measuring and transportation environment. Ni was, however, a less probable contaminant than other transition metals such as Fe, Cu and Zn in semiconductor manufacturing. The UCS also carried out round robins of both TXRF and AAS for these metals, but the results were not satisfactory: the analytical results using AAS, the average of which was planned to use for the calibration of TXRF instrument, did not agree well among the laboratories which participated, so that the calibration method of TXRF instrument was not established. The accuracy of quantitative analysis using TXRF was not determined for these metals by the work of the UCS. Laboratorie...