Spin polarised electron energy loss spectroscopy (SPEELS) is an experimental tech- nique that can resolve spin dependent information about the target. The experiment is the same as an EELS measurement with the distinct difference of having either an incident electron beam that is polarised [156, 157], or a spin polarised detection system [158], or in some cases both [159, 160]. The polarised incident electron beam gives rise to spin specic transitions at a low energy loss [161]. The spin dependent processes can be di- vided into two types of processes, those that ip the incident electron's spin (ip) and those that do not ip the incident electron's spin (non-ip). Flip scattering processes (Fig. 6.6: a) involve the incident electron occupying an empty state above the Fermi level and transferring its energy to a bound electron with opposite spin, which is ejected, creat-
ing an electron-hole pair (Stoner excitation) [161]. Non-ip processes (Fig. 6.6: b) can involve both direct or exchange scattering. Direct, non-ip scattering occurs via dipole scattering (when the incident electron has long-range electric eld interactions with the electrons in the target) or impact scattering (when the incident electron penetrates into the sample and is directly scattered). Exchange scattering also occurs via impact scattering. The combination of these different types of spin dependent inelastic scattering produces small features in the SPEELS spectrum. The cross sections of these spin dependent in- elastic scattering processes is many orders smaller than plasmon excitations and thus are most noticeable at very small energy loses where the elastic peak no longer dominates and before the rst plasmon excitation is too intense.
E
F 1 2 3a)
b)
E
F 1 2Figure 6.6: a) The spin-ip (Stoner excitation) scattering process broken into three stages. The
incident↑electron occupies an empty orbital (1), couples with a↓electron below the Fermi level (2), ejecting the↓electron. b) A Non-ip scattering process, shows the long range electric eld
uctuations that can eject a bound electron of the same spin orientation, whilst the incident electron is still well outside the target.
Previous studies into inelastic spin dependent features in SPEELS spectrums [162, 163, 164, 165] have shown quite broad and featureless spectra. The eld of SPEELS became stagnant in the mid 90's, and was maligned to a eld for theorists to study. In 2006 a study by Komesu et. al. [166] on the SPEELS spectra on magnetic Fe showed signs of peaks associated with Stoner excitations in both the minority and majority spin spectra.
The experiment of Komesu et. al. [166] has the same scattering angle (45 ◦) and a slightly worse energy resolution (0.6 eV, compared to 0.5 eV) than the EMS spectrometer.
1 1.5 2 2.5 3 Spin Majority
Spin Majority spectra
Spin Integrated spectra
In te n si ty (a rb . u n it s)
Energy Loss (eV) Theoretical peak
(2.3 eV)
d)
Experimental Peak (2.35 eV)
"Guide to the eye"
1 1.5 2 2.5 3
Spin Minority Spin Minority spectra Spin Integrated spectra
In te n si ty (a rb . u n it s)
Energy Loss (eV) Experimental Peak
(2.05 eV)
e)
"Guide to the eye"
Theoretical peak (1.9 eV)
Figure 6.7: The spin integrated (a), spin majority (b) and spin minority (c) SPEELS spectra of
Komesu et. al. for theΓ-H direction of magnetised Fe. The spin majority and spin minority
spectra which were measured by the left/right scattering from the Mott polarimeter have been re- scaled to the same height as the spin integrated spectra, hence the larger error bars. The results of this study are presented in the same fashion with the spin majority (d) and spin minority (e). The calculated Stoner excitations and direct transitions [167] are shown in the spin majority (d) and spin minority (e) spectra for this study.
The main differences were the experiment of Komesu et. al. was performed at 300 eV, making his experiment more sensitive to Stoner excitations and direct transitions. The ANU experiment required much higher energy (25 keV) as it was performed in transmis- sion mode. Another difference is that Komesu had a Mott polarimeter after the interaction region so they could separate out the spin contributions after the interaction region, whilst the EMS spectrometer has a spin polarised electron source (21 %). In an effort to sup-
port the data of Komesu et. al. which was performed on aΓ-H magnetised Fe sample, a SPEELS experiment in the same direction of magnetised Fe (h100i,Γ-H) was performed
in the EMS spectrometer. In the EMS measurements the spin-ip (Stoner excitations) are initiated by spin-down incident electrons, in keeping with the nomenclature of Komesu this spectra is labelled as majority spectra. Alternatively, direct transitions are initiated by incident electrons that are spin-up, that EMS spectra was labelled as minority spec- tra. The two data sets are both presented in raw EELS spectra, as an spin integrated EELS spectra and spin-up and spin-down EELS spectra, they are compared in Fig. 6.7.
The experimental results of Komesu et. al. for the minority spin state (Fig. 6.7: c) has a feature measured at 2.1 eV which agrees quite well with the calculated feature at 2.3 eV [167]. The calculated 2.3 eV feature corresponds to a Stoner excitation between the unoccupied∆↓2 and the occupied ∆↑2 states, which would appear in the minority spin states spectra due to the spin-ip interaction. The majority spin state spectra (Fig. 6.7: b) of Komesu et. al. shows two features near 1.8 eV and 2.6 eV. Komesu et. al. believe that those features are due to direct transitions which is a non-ip process. Photoemission spectra [168] has measured band positions that would predict a direct transition peak of 2.8 eV, that would agree well with the 2.6 eV feature (Fig. 6.7: b).
The experimental results of this study of the spin-polarised electron-energy-loss spec- trum measured with the spin vectors of the incident and bound electrons along the Γ-H (h100i) direction in magnetised Fe produces some results which are in mild agreement to
that of Komesu et. al.. The majority spectra (Fig. 6.7: d) which shows spin-ip excita- tions such as Stoner excitations, has one obvious feature at 2.35 eV, which compares well to the theoretical value (2.3 eV) for the∆↓2−∆↑2Stoner excitation. This Stoner transition was measured at 2.5 eV by Komesu et. al., and the two energy levels were measured at 2.8 eV apart by photoemission data [168]. It is plausible that there maybe a peak below 2 eV, that would correspond with the measurement of Komesu (1.8 eV) and could be due to the∆↓5−∆↑5Stoner transition that has a theoretical value of 2.1 eV [167]. The validity of this second feature is dubious at best.
The minority spectra from this study (Fig. 6.7: e) corresponds to direct transitions, that is transitions that involve the incident spin-up electron colliding with a spin-down bound electron. One obvious feature (2.05 eV) that may be attributed to a direct∆↓2−∆↓5
excitation. Calculations by Callaway and Wang [167] show this feature to be at 1.9 eV, and the study of Komesu et. al. [166] show this feature to be at 2.0 eV.
Poor statistical resolution and less sensitivity raises the question whether these fea- tures are actual Stoner excitations or are from a combination of statistical inaccuracies, incomplete incident polarisation or other inelastic processes. The position of the features in both this study and the study by Komesu et. al. [166] do agree well with theory and other experimental measurements [167, 168]. Due to the poor statistical accuracy and clarity of these features especially in this study but also in the study of Komesu et. al., it is not entirely convincing as to whether these peaks can be decisively attributed to Stoner excitations and direct transitions.