Sample Preparation for Electron Probe Microanalysis—Pushing the Limits

There are two fundamental considerations in preparing samples for electron probe microanalysis (EPMA). The first one may seem obvious, but we often find it is overlooked. That is, the sample analyzed should be representative of the population from which it comes. The second is a direct result of the assumptions in the calculations used to convert x-ray intensity ratios, between the sample and standard, to concentrations. Samples originate from a wide range of sources. During their journey to being excited under the electron beam for the production of x rays there are many possibilities for sample alteration. Handling can contaminate samples by adding extraneous matter. In preparation, the various abrasives used in sizing the sample by sawing, grinding and polishing can embed themselves. The most accurate composition of a contaminated sample is, at best, not representative of the original sample; it is misleading. Our laboratory performs EPMA analysis on customer submitted samples and prepares over 250 different calibration standards including pure elements, compounds, alloys, glasses and minerals. This large variety of samples does not lend itself to mass production techniques, including automatic polishing. Our manual preparation techniques are designed individually for each sample. The use of automated preparation equipment does not lend itself to this environment, and is not included in this manuscript. The final step in quantitative electron probe microanalysis is the conversion of x-ray intensities ratios, known as the "k-ratios," to composition (in mass fraction or atomic percent) and/or film thickness. Of the many assumptions made in the ZAF (where these letters stand for atomic number, absorption and fluorescence) corrections the localized geometry between the sample and electron beam, or takeoff angle, must be accurately known. Small angular errors can lead to significant errors in the final results. The sample preparation technique then becomes very important, and, under certain conditions, may even be the limiting factor in the analytical uncertainty budget. This paper considers preparing samples to get known geometries. It will not address the analysis of samples with irregular, unprepared surfaces or unknown geometries.