Interpretation of spin-polarized electron energy loss spectra

We study the origin of the structure in the spin-polarized electron energy loss spectroscopy (SPEELS) spectra of ferromagnetic crystals. Our study is based on a 3d tight-binding Fe model, with constant onsite Coulomb repulsion U between electrons of opposite spin. We find it is not the total density of Stoner states as a function of energy loss which determines the response of the system in the Stoner region, as usually thought, but the densities of Stoner states for only a few interband transitions. Which transitions are important depends ultimately on how strongly umklapp processes couple the corresponding bands. This allows us to show, in particular, that the Stoner peak in SPEELS spectra does not necessarily indicate the value of the exchange splitting energy. Thus, the common assumption that this peak allows us to estimate the magnetic moment through its correlation with exchange splitting should be reconsidered, both in bulk and surface studies. Furthermore, we are able to show that the above mechanism is one of the main causes for the typical broadness of experimental spectra. Finally, our model predicts that optical spin waves should be excited in SPEELS experiments.