The simulated line profiles, from MARX, are fit with a Gaussian plus Lorentzian profile.
where r
is the peak of the Gaussian, 
is the Gaussian
width and r is the distance from the peak. For the Lorentzian
component 
is the Lorentzian width, r
is the
Lorentzian center not necessarily the same as the Gaussian peak, and
a-0 is the relative normalization with respect to the Gaussian.
Figure 6 shows the variation of the Gaussian 
and relative Lorentzian normalization as a function of wavelength
while figures 7 and 8 shows the same
for the HEG and LEG spectra.
Figure 6: MEG Gaussian Width and Lorentzian amplitude vs wavelength
Figure 7: HEG Gaussian Width and Lorentzian amplitude vs wavelength
Figure 8: LEG Gaussian Width and Lorentzian amplitude vs wavelength
In these fits the Gaussian peak is set to 1.0 and the Lorentzian 
is fixed at 0.012 Å.
As can be seen in figure 6 the Gaussian width between
about 10 and 40 Å rises from a low of 
0.007 to a high of about
0.0105 Å . The right panel shows the Lorenzian amplitude (relative
to the Gaussian) as a function of wavelength. At long wavelengths the
Lorentzian is a weak component but at short wavelengths the Lorentzian
component dominates over the Gaussian component. This effect is
expected and is mostly due to mirror scattering (see POG figure 4.6)
and so the scattered component, represented by the Lorentzian, dominates
at short wavelength (high energy).