Principal component analysis (PCA), microbeam X-ray Fluorescence Spectrometry (µXRF), and Monte Carlo simulations are used to illustrate a methodology for assessing the sample microheterogeneity for large sets of materials. PCA is demonstrated to determine a minimum sample size that can be used without concern that sample heterogeneity is contributing significantly to measurement uncertainty. The amount of data required for an accurate PCA model is discussed employing both empirical data and simulated data generated using a Monte Carlo approach to illustrate that sampling methodology is critical. Random data collection is compared to rastering across a sample as typically done with microanalytical techniques. The use of random data collection is shown to reduce analysis time by one order of magnitude for some samples without significantly reducing the quality of data for estimating minimum sample size. Examples are shown for SRM 1635a Trace Elements in Coal (Subbituminous) and SRM 1729 Tin Alloy (97Sn-3Pb) with estimated minimum sample sizes of 2 mg and 640 µg, respectively.
Citation: X-Ray Spectrometry
Pub Type: Journals
µXRF, microheterogeneity, principal component analysis (PCA), reference materials, elemental analysis, sampling