Insufficient absorption correction at low kV?

[dtlalexander]

Working on quantification of a "standard" sample of DyScO3. Using the actual experimental values of tilt_stage=0.0, elevation_angle=18.0, azimuth_angle=45.0 giving take_off_angle=18.0, the percentage of O in the sample is greatly underestimated at 37.0 at.%. Also, the fitting of the background is not good at low energy (Sc L peak, O K peak), being too high especially under Sc L peak.

By "tricking" the code using a very low take_off_angle, the estimated percentage of O is much closer to reality. For instance, using: elevation_angle=18.0, azimuth_angle=315.0 to give take_off_angle=2.7, the percentage of O in the sample is estimated at 60.2 at.%. Also, the low energy background is suppressed, and fitting of the Sc L peak is much better. Note that the output is very sensitive to take_off_angle at such low angles. Also note that X-ray lines were split at 3.0 keV for Sc, Cu (from the grid) and Dy.

[adriente]

Since the 'strength' of the absorption scales linearly with the cosec of the take-off angle, the density and the thickness the culprit is between those three. Another culprit could be the absorption coefficient that is based on theoretical values imported from exspy (hyperspy) database. However, the fact that the Sc L is better fitted with a lower take-off angle seems to point at the "strength" of the absorption and not the shape of the absorption coefficient.

When using the low take-off angle, is the Sc quantification yielding better results ?

Do you have a good estimates of the experimental density and thickness of your sample ?