Is root N a sufficient measure of the standard uncertainty in X-ray spectroscopy?

摘要

There is a magnitude larger scatter in the experimental data of fundamental parameters than the claimed error estimate. We give examples from recent compilations of excitation and decay parameter values for the untenable large scatters, indicating methodological problems. One is the improper use of uncertainty estimation. The measured spectrum is not expected to follow Poisson distribution. We report proper statistical uncertainty calculations. It implies a two to five times larger uncertainty but still does not account for the large scatter. The other possible explanation could be rooted in the ill-posed problem of exponential analysis, as radiation measurement belongs to this category. We give evidence from particle-induced X-ray emission and X-ray fluorescence for additional exponential terms, thus leading to multi-exponential analysis. This could explain the large scatter, as the usual square root of counts rule cannot be used for the standard uncertainty. We present a novel approach where discriminators are used to reduce the number of exponentials and the discriminated events are also processed and collected into a separate spectrum. Analyzing both spectra and the live time and dead time clocks allows the determination of the true input counts. It is a non-extended dead time approach. With this approach, we have a much reduced statistical uncertainty, and both the total spectrum and the fractional spectrum have the same uncertainty. As an independent quality assurance tool, the time interval histogram analysis is also presented. Copyright (C) 2017 John Wiley & Sons, Ltd.