Electron-probe microanalysis of uranium and uranium alloys poses several problems, such as rapid oxidation, large poorly constrained correction factors, and a large number of characteristic x-ray lines. We show that U-metal can grow 10 nm of oxide within ~20 s of air exposure, increasing to 15–20 nm within a few minutes, which can produce a 30% quantification error at 5 kV. A 15 nm carbon coating on the UO2 reference material also produces an ~30% quantification error of the uncoated but surface oxidized U sample at 5 kV. Correcting for both the coating and oxide improved the analysis accuracy to better than ±1% down to 7 kV and ~2% at 5 kV, but the error increases strongly below this. The measurement of C in U identified a previously unreported U N6–O4 line interference on the C Kα peak, which can produce over 1% error in the analysis total. Oxide stoichiometry was demonstrated to have only a small impact on quantification. The measurement of the O Kα and U Mα mass absorption coefficients in U as 9,528 and 798 cm2/g, respectively, shows good agreement with recently published values and also produces small differences in a quantification error.

Low voltage electron probe microanalysis (EPMA) of metals can be complicated by the presence of a surface oxide. If a conductive coating is applied, analysis becomes one of a three-layer structure. A method is presented which allows for the coating and oxide thicknesses and the substrate intensities to be determined. By restricting the range of coating and oxide thicknesses, tc and to respectively, x-ray intensities can be parameterized using a combination of linear functions of tc and to. tc can be determined from the coating element k-ratio independently of the oxide thickness. to can then be derived from the O k-ratio and tc. From tc and to the intensity components of the k-ratios from the oxide layer and substrate can each be derived. Modeled results are presented for an Ag on Bi2O3 on Bi system, with tc and to each ranging from 5 to 20 nm, for voltages of 5–20 kV. The method is tested against experimental measurements of Ag- or C-coated samples of polished Bi samples which have been allowed to naturally oxidize. Oxide thicknesses determined both before and after coating with Ag or C are consistent. Predicted Bi Mα k-ratios also show good agreement with EPMA-measured values.

Secondary fluorescence (SF), typically a minor error in routine electron probe microanalysis (EPMA), may not be negligible when performing high precision trace element analyses in multiphase samples. Other factors, notably wavelength dispersive spectrometer defocusing, may introduce analytical artifacts. To explore these issues, we measured EPMA transects across two material couples chosen for their high fluorescence yield. We measured transects away from the fluorescent phase, and at various orientations with respect to the spectrometer focal line. Compared to calculations using both the Monte Carlo simulation code PENEPMA and the semi-analytical model FANAL, both codes estimate the magnitude of SF, but accurate correction requires knowledge of the position of the spectrometer with respect to the couple interface. Positioned over the fluorescent phase or otherwise results in a factor of 1.2–1.8 of apparent change in SF yield. SF and spectrometer defocusing may introduce systematic errors into trace element analyses, both may be adequately accounted for by modeling. Of the two, however, SF is the dominant error, resulting in 0.1 wt% Zn apparently present in Al at 100 μm away from the Zn boundary in an Al/Zn couple. Of this, around 200 ppm Zn can be attributed to spectrometer defocusing.

Electron beam-induced carbon contamination is a balance between simultaneous deposition and erosion processes. Net erosion rates for a 25 nA 3 kV beam can reduce a 5 nm C coating by 20% in 60 s. Measurements were made on C-coated Bi substrates, with coating thicknesses of 5–20 nm, over a range of analysis conditions. Erosion showed a step-like increase with increasing electron flux density. Both the erosion rate and its rate of change increase with decreasing accelerating voltage. As the flux density decreases the rate of change increases more rapidly with decreasing voltage. Time-dependent intensity (TDI) measurements can be used to correct for errors, in both coating and substrate quantifications, resulting from carbon erosion. Uncorrected analyses showed increasing errors in coating thickness with decreasing accelerating voltage. Although the erosion rate was found to be independent of coating thickness this produces an increasing absolute error with decreasing starting thickness, ranging from 1.5% for a 20 nm C coating on Bi at 15 kV to 14% for a 5 nm coating at 3 kV. Errors in Bi Mα measurement are <1% at 5 kV or above but increase rapidly below this, both with decreasing voltage and increasing coating thickness to 20% for a 20 nm coated sample at 3 kV.

The accuracy to which Cu and Al coatings can be determined, and the effect this has on the quantification of the substrate, is investigated. Cu and Al coatings of nominally 5, 10, 15, and 20 nm were sputter coated onto polished Bi using two configurations of coater: One with the film thickness monitor (FTM) sensor colocated with the samples, and one where the sensor is located to one side. The FTM thicknesses are compared against those calculated from measured Cu Lα and Al Kα k-ratios using PENEPMA, GMRFilm, and DTSA-II. Selected samples were also cross-sectioned using focused ion beam. Both systems produced repeatable coatings, the thickest coating being approximately four times the thinnest coating. The side-located FTM sensor indicated thicknesses less than half those of the software modeled results, propagating on to 70% errors in substrate quantification at 5 kV. The colocated FTM sensor produced errors in film thickness and substrate quantification of 10–20%. Over the range of film thicknesses and accelerating voltages modeled both the substrate and coating k-ratios can be approximated by linear trends as functions of film thickness. The Al films were found to have a reduced density of ~2 g/cm2.

Automated phase maps are an important tool for characterizing samples but the data quality must be evaluated. Common options include the overlay of phases on backscattered electron (BSE) images and phase composition averages and standard deviations. Both these methods have major limitations. We propose two methods of evaluation involving principal component analysis. First, a red–green–blue composite image of the first three principal components, which comprise the majority of the chemical variation, which provides a good reference against which phase maps can be compared. Advantages over a BSE image include discriminating between similar mean atomic number phases and sensitivity across the entire range of mean atomic numbers present in a sample. Second, principal component maps for identified phases, to examine for chemical variation within phases. This ensures the identification of unclassified phases and provides the analyst with information regarding the chemical heterogeneity of phases (e.g., chemical zoning within a mineral or mineral chemistry changing across an alteration zone). Spatial information permits a good understanding of heterogeneity within a phase and allows analytical artifacts to be easily identified. These methods of evaluation were tested on a complex geological sample. K-means clustering and K-nearest neighbor algorithms were used for phase classification, with the evaluation methods demonstrating their limitations.

Quantification of first series transition metal Lα X-rays is hampered by absorption and in some cases transition probabilities (fluorescence yields) varying with chemical bonding. Compound mass absorption coefficients for Fe Lα were measured in the olivine solid solution series [Forsterite (Mg2SiO4) to Fayalite (Fe2SiO4)] and the mass absorption coefficients for Fe Lα absorbed by Fe were calculated. The mass absorption coefficients vary systematically between Fo83 and Fo0. Using the measured mass absorption coefficients for both standard and unknown and by correcting for a systematic discrepancy, consistent with varying partial fluorescence yields, a good agreement between calculated k-ratios and measured k-ratios is achieved. The systematic variations allow quantification of unknown k-ratios. The described method of quantification requires modification of matrix correction routines to allow standards and unknowns to have different mass absorption coefficients, and to incorporate solid solution mass absorption coefficients and partial fluorescence yield corrections derived from regression of experimental data.

A prototype Peltier thermoelectric cooling unit has been constructed to cool a cold finger on an electron microprobe. The Peltier unit was tested at 15 and 96 W, achieving cold finger temperatures of −10 and −27°C, respectively. The Peltier unit did not adversely affect the analytical stability of the instrument. Heat conduction between the Peltier unit mounted outside the vacuum and the cold finger was found to be very efficient. Under Peltier cooling, the vacuum improvement associated with water vapor deposition was not achieved; this has the advantage of avoiding severe degradation of the vacuum observed when warming up a cold finger from liquid nitrogen (LN2) temperatures. Carbon contamination rates were reduced as cooling commenced; by −27°C contamination rates were found to be comparable with LN2-cooled devices. Peltier cooling, therefore, provides a viable alternative to LN2-cooled cold fingers, with few of their associated disadvantages.

The effect of carbon contamination on the analysis of carbon-coated silicate minerals at 5 kV for X-ray energies 0.7–4 keV is examined. For individual spot analyses, carbon is found to deposit adjacent to the beam spot forming ring-shaped deposits with no impact on the analysis. Carbon contamination becomes important for closely spaced analyses such as multipoint transects, where each subsequent analysis overlaps the carbon ring of the previous analysis. X-ray intensity loss due to contamination is most severe for low-overvoltage elements such as Ca K consistent with carbon deposition effectively reducing beam energy. Rates of contamination are calculated and the use of a liquid nitrogen cold trap is shown to greatly reduce the amount of carbon deposited. A complimentary empirical correction is developed to correct for X-ray intensity loss from measured carbon, assuming the carbon is a film, and is compared with corrections derived from thin film calculations. PENELOPE electron probe microanalysis (PENEPMA) calculations confirm that asymmetry of the carbon deposition can be ignored for X-ray energies where intensity loss is predominantly through energy loss of beam electrons. Using a cold trap and/or an empirical correction high spatial resolution analysis (ca. 400 nm between points) is achievable with analytical errors of ca. 1–3%.