Propagation of Errors in the Precision and Accuracy of Quantitative Electron Microbeam Analysis

Abstract

A common question, but one typically difficult to answer, is “How accurate is a microprobe or SEM analysis?” Commercial software for most EDS and WDS systems print out error limits that are typically calculated standard deviations using Gaussian counting statistics based on the accumulated peak and background counts. This is not a true measure of even the measurement precision, since random instrumental instabilities typically degrade “true” measurement precision (as determined by replicate measurements on standards) to 1.5 - 3+ times that predicted from Gaussian statistics. in addition to these random instrumental variations are a plethora of possible systematic and semirandom errors such as detector or faraday cup non-linearities, errors in estimation of background, EDS spectral deconvolution errors, WDS positional non-reproducibility, temperature and pressure variation effects on spectrometer efficiencies, high voltage drift or sample charging, surface contamination or oxidation, and sample roughness or tilt.