Polarized Cross Sections Order Error¶
In versions of Refl1D before 1.0.0, there was an error in the interpretation of the outputs of the magnetic calculation kernel. We were inadvertently using an opposite convention for the direction of the Aguide angle than what is used for the EPS angle in the Polarized Neutron Reflectometry (PNR) book chapter by Majkrzak et al. [1], while simultaneously using the outputs of the magnetic calculation kernel in reversed order.
Choosing an opposite convention for the Aguide angle causes the outputs of the calculation to be reversed when there is no splitting in the spin-flip cross sections, so these two choices effectively cancel each other out almost all the time.
The exception is when there is significant splitting in the spin-flip cross sections, which occurs when there is a significant applied field and a significant projection of the magnetic moment perpendicular to that applied field. In this case, the energy corrections are applied to the wrong elements, leading to incorrect results.
We have corrected this error in the release 1.0.0 of Refl1D.
Note that as a result of the original error in interpreting the outputs of the Refl1D magnetic kernel, the reflectivity figures in the original Journal of Applied Crystallography article on PNR [2] in large fields have the cross sections labeled in reverse order.
Backward Compatibility¶
To maintain backward compatibility, the Aguide parameter in the model is now interpreted as \(\text{Aguide} \equiv -\text{EPS}\), where \(\text{EPS}\) is the angle between the sample normal (z-axis) and the applied guide field, as shown in figure 1.114 in the book chapter.
The output of the magnetic kernel is in the order \(R^{++}, R^{+-}, R^{-+}, R^{--}\), but the output of refl1d.sample.reflectivity.magnetic_amplitude is reversed to match the order of the cross sections as stored in refl1d.probe.PolarizedNeutronProbe: \(R^{--}, R^{-+}, R^{+-}, R^{++}\).
For existing models, unless there is splitting in the spin-flip cross sections, the calculated results will remain the same as before.
Validation¶
See Gepore Validation example for a comparison of the results including the correction.