Surface irregularities – We have discussed this already, in a section, above.
Etched samples – Samples that have been chemically etched or etched by a plasma beam/cleaner are commonly rougher than unetched samples, which leads to anomalous diffraction. This type of roughness is induced and controllable to a certain extent, which is why etching qualifies as a systematic error.
Polish – In order to minimize surface effects, samples should be polished with a diamond, alumina or silica abrasive with a grain-size of < 1 μm.
Specimen homogeneity – We have already discussed the effects of having more than one phase present in the excitation volume, above.
Non-orthogonal geometry – As sometimes happens with grain-mounts that are made into an epoxy ‘puck’ format, and especially for those that are automatically polished, the surface of the mounts tends to slope in one direction, i.e, the surface is not perpendicular to the central axis of the puck. When these pucks are placed in ring holders, they almost always tilt as the screw is tightened, and the surface of the sample ends up inclined at an angle to the horizontal plane. X-rays emerging from the tilted sample won't proceed towards the diffractor crystal properly, as the Bragg condition is violated by the angle of tilting. Therefore, X-ray intensities will be reduced relative to a non-inclined, ideal geometry, giving rise to errors in the accuracy of the analysis.
Edge effects – Materials mounted in epoxy and then polished with loose abrasives commonly develop channels or grooves in the epoxy at the sample-epoxy interface, which leads to rounding of the sample edge. The rounding of the sample results in a reduced absorption path-length of the beam from ideal/assumed, resulting in an erroneous analysis.
Oxide/oxidation coating – For samples and standards that readily oxidize during storage (before analysis), such as Ti, Al, Mg, Mn, an oxide coating can reach a thickness of fractions of a micron, and result in a reduced intensity being collected for a standard – this results in an over-estimate of the amount of the appropriate element in the unknown, as the denominator of the K-ratio will be reduced below what it would be if the standard was not oxidized.
Polishing artifacts – Polishing media can become trapped in pores in the sample surface, or embedded in samples that are considerably softer than the polishing media. I have seen this with 0.25-micron diamond particles from a polish becoming stuck in the surface of Sn-metal. Also, if Pb polishing laps are used to polish the sample, Pb metal from the lap can become trapped in pores or along grain boundaries – this leads to a nightmare if one is analyzing zircons or monazites, because the Pb from the lap may smear across the surface of a sample, giving rise to ‘false’ X-ray intensities for Pb, and potential reduced counts for lighter elements whose counts get absorbed by the Pb.
Smear coating – Smear coating refers to the smearing of softer minerals across the surfaces of adjacent ones during the polishing process, causing contamination. This can happen with Sn and Pb in certain samples. This effect is more relevant for trace-element analysis than for major elements.
Carbon-coat thickness – Ideally, to remove yet another variable/source of error from our analyses, it is best if the thickness of the carbon-coat is the same for our standards as it is for our unknowns. Kerrick et al. (1973) discovered that this effect of carbon-coating thickness was worse for light elements, and that in the case of fluorine, a difference in coating thickness of about 200 Å between the sample and standard resulted in a 4 % difference in X-ray intensity for the F Kα X-ray (see next page for illustration from Kerrick’s 1973 paper).
