Resonant Auger spectromicroscopy in ultrathin Fe films on W(110)

L3M 2,3 M 2,3 Auger transition is measured near the L3 resonance of ferromagnetic Fe films on W(110). The kinetic energies of the Auger peaks display the typical Raman behaviour for photon energies well below the absorption threshold, where the Auger energy follows the changes in the photon energy. Classical Auger behaviour with constant kinetic energy sets in at about 1.5 eV below the L3 resonance independently from the number of Fe layers down to the monolayer thickness. Strong x-ray circular magnetic dichroism is observed at the L3 edge in the entire L3M 2,3 M 2,3 Auger spectrum. Different Auger features originating from the final state with two 3p core holes show slight variations in the dichroic signal, which is attributed to the exchange interaction between the core holes and the valence band. Finally, XMCD-PEEM magnetic domain imaging using Auger electrons is demonstrated with a high level of contrast and lateral resolution approaching that of imaging with secondary photoelectrons.


Introduction
Auger electron spectroscopy (AES) is known to be a wellestablished surface chemical analysis tool using both electrons and x-rays as excitation source [1]. Remarkably, tunable x-ray energy and polarization available at synchrotron x-ray sources allowed AES to probe electronic effects beyond simple chemical characterization [2]. Resonant AES has attracted attention as a probe of femtosecond electron dynamics using the core-hole lifetime as an internal clock [3][4][5][6][7]. In particular, near absorption resonances, the kinetic energy of Auger peaks * Authors to whom any correspondence should be addressed.
Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
shows Raman behavior due to the competition between corehole decay and relaxation of the core-excited state, which gives information on the relaxation dynamics [8,9].
The application of resonant AES has been mostly limited to bulk materials or weakly-interacting physisorbed adsorbate layers. Ultrathin metal films and especially spatially-resolved measurements have been largely ignored due to the stringent requirements on the signal levels. There have been few cases in which x-ray magnetic circular dichroism photoemission electron microscopy (XMCD-PEEM) was applied to imaging magnetic domains using Auger electrons [10][11][12]. Overall, magnetic imaging with Auger electrons has not been pursued intensively, likely due to the low count rates deemed insufficient for spatially-resolved measurements. On the other hand, the motivation of imaging buried layers, which should be possible due to the larger inelastic mean free path of electrons at high kinetic energy [13], is still valid, and the everincreasing brightness of synchrotron sources circumvents the issue related to image statistics. x-ray absorption spectrum of Fe/W(110) measured using secondary electron yield and linearly-polarized x-rays. (b) Fe L 3 M 2,3 M 2,3 Auger spectrum measured at 708.1 eV photon energy, which corresponds to the peak of the L 3 resonance. At the highest kinetic energy, the tail of the Fe 3s photoemission peak can be seen.
Along these lines, we apply resonant L 3 M 2,3 M 2,3 AES measurements to ultrathin Fe films grown on W(110). The LMM Auger transition was chosen, in particular, to avoid the presence of direct photoemission channels at the resonance condition. By tuning the photon energy near the L 3 x-ray absorption threshold of Fe, we investigate the photon-energy dependent Auger kinetic energy dispersion as a function of the Fe film thickness. The transition from Raman regime to classical Auger behaviour is observed at photon energies at more than 1 eV below the L 3 resonance threshold, indicating excited state dynamics at a time scale below 1 fs. Then, XMCD spectroscopy and imaging measurements are performed in a comparative manner using both secondary photoelectrons and L 3 M 2,3 M 2,3 Auger electrons. In spite of the low count rates, we demonstrate discreet image quality and a lower background in the case of Auger electrons. Lastly, the details of the XMCD spectrum measured with the Auger electrons slightly deviate from the dichroism in total yield measurements, an observation which can be attributed to the exchange interaction between the core holes in the Auger final state and the spinpolarized valence band.

Experimental method
Film growth, characterization and x-ray spectroscopy measurements were all carried out in the Spectroscopic PhotoEmission and Low Energy Electron Microscope (SPELEEM) at the Nanospectroscopy beamline, Elettra synchrotron [14,15]. The undulator source available at the beamline can provide both circularly and linearly polarized light in the soft x-ray regime from 30 eV up to 1300 eV. The photon energy resolution at the Fe L-edge (at around 700 eV) reaches about 0.2 eV. The photoemitted electrons are collected along the surface normal by the microscope imaging system, accelerated to high electron energies, filtered in energy using a hemispherical energy analyzer, and imaged on a multichannel plate detector. In order to collect Auger spectra as a function of photon energy in an efficient manner, the dispersive plane of the energy analyzer was imaged onto the detector, giving a photoemission spectrum within a 12 eV window.
The x-ray beam was incident at 16 • grazing angle from the sample surface, along the Fe [110] direction. Most of the Auger study was carried out using linear p-polarization (close to the surface normal direction), except for the x-ray magnetic circular dichroism measurements obtained by switching the helicity of circular polarization. The x-ray energy was calibrated using the energy distance between the first and second harmonics of the variable line grating monochromator.
Ultrathin Fe films were grown on a W(110) singlecrystal surface using an electron-beam evaporator installed within the SPELEEM instrument. W(110) was cleaned using a standard procedure based on prolonged annealing in 10 −6 mbar molecular oxygen at about 1350 K followed by high-temperature flashes to about 2300 K in ultrahigh vacuum. Throughout the growth and measurements, the pressure in the ultrahigh vacuum vessel remained at low 10 −10 mbar. The evaporation rate was calibrated from the completion of the pseudomorphic Fe monolayer (ML) in low-energy electron microscopy (LEEM) mode. The thickness of Fe films was determined by monitoring the number of oscillations in the energy-dependent electron reflectivity curves [16]. All measurements were done at 300 K.

Thickness-dependent resonant L 3 M 2,3 M 2,3 Auger emission in ultrathin Fe films
The x-ray absorption spectrum of a 12 ML thick Fe film on W(110) is shown in figure 1(a). The typical metallic Fe XAS curve features both the L 3 and L 2 peaks. Above the L 3 resonance, the L 3 M 2,3 M 2,3 (Fe 2p-3p3p) Auger transition takes place generating a two 3p core-level hole final state. Figure 1(b) shows this transition at the L 3 peak, i.e. 708.1 eV photon energy. The two-hole final state of this transition is split by exchange interaction between the 3p holes into a triplet ( 3 P) and two singlet ( 1 S, 1 D) states, which are labeled on figure 1(b) [17]. Importantly, the absence of a direct photoemission peak near the L 3 M 2,3 M 2,3 Auger lines simplifies the following analysis.
Resonant Auger allows probing hybridization and localization of photo-excited electrons. The difference between the photon energy at which Raman and Auger behaviors cross and the energy of the resonance contains information not only on the electronic structure but also on the dynamics of the electrons, as a consequence of the photon-created core-hole [18,19]. In order to identify the transition from Raman to classical Auger behavior in the L 3 M 2,3 M 2,3 Auger level, we monitor the kinetic energy of the triplet peak 3 P, which appears sharper and more pronounced as compared to the singlet peaks. The result is given in figure 2 as a function of Fe thickness from 1 ML up to 5 ML. The Raman-Auger transition can be clearly observed for all thicknesses as a shift from a photon energydependent peak position to a nearly constant kinetic energy as expected from a classical Auger peak. The three main observations are: (i) the transition takes place more than 1 eV before the L 3 threshold, (ii) the slope of the Auger energy versus photon energy dispersion is close to, but slightly above, 1, in the Raman regime, (iii) the transition point is sharper for the Fe ML and bilayer, whereas it has a more rounded shape for the thicker layers.
The Raman-Auger transition energy for bulk Fe, as well as Cr and Ni, was reported for the LVV Auger lines [18]. In qualitative agreement with the bulk Fe case, the classical Auger behaviour (i.e. constant Auger energy) sets in at more than 1 eV below the maximum of the L 3 resonance also in the case of the LMM transition, here reported. In previous studies on manganite samples, the onset of constant Auger kinetic energy regime was attributed to the dependence of the effective scattering time on the detuning from the photo-absorption resonance energy [8]. The idea is based on the scattering duration, τ , which depends on the decay of the excited state due to the core-hole lifetime broadening (Γ) and its oscillatory behaviour due to the detuning from the exact transition energy (Ω) [9]: This means that the scattering effectively becomes faster as the photon energy is detuned from the resonance. The probability of deexcitation into the Auger final state before the original photoelectron can dissipate its energy is higher for larger detuning values. Thus, the Raman-Auger transition was suggested to take place in solid samples when the detuningdependent τ becomes shorter than the charge transfer time [8].
Raman-Auger transition is seen at Ω = 1.5 eV for ML Fe, and in the range from 1.3 eV to 1.5 eV for the thicker films. The exact position of the transition is difficult to identify for the thicker samples as the change in the Auger peak position follows a rounded curve without a well-defined kink. Nevertheless, we can conclude that the relaxation dynamics of the core-excited state before the Auger decay happens at a time scale similar for all Fe thicknesses, in spite of the different crystalline and chemical environment, in particular, for the Fe ML. Taking the core-hole lifetime broadening as Γ = 0.5 eV [20], according to equation (1) this relaxation time scale is found to be about 0.4 ± 0.1 fs.

X-ray magnetic circular dichroism at L 3 M 2,3 M 2,3
X-ray magnetic circular dichroism in LMV Auger emission had been studied both in spectroscopy and imaging modes in the case of Fe [11]. Here, we investigate the circular dichroism for the L 3 M 2,3 M 2,3 line for ultrathin Fe films on W(110).   by collecting the secondary photoelectrons. The corresponding XMCD spectra obtained by collecting Auger electrons traced to the singlet ( 1 D at 588 eV) and triplet ( 3 P at 595 eV) peaks can be observed in figures 3(b) and (c), respectively. Most importantly, both Auger features show strong circular dichroism with the same sign and, not surprisingly, the entire Auger line follows the polarization-dependent x-ray absorption cross-section at the resonance.
Even though the asymmetry on the white line intensity is nearly identical for all three spectra, the baseline intensity before the resonance is dramatically reduced for the Auger lines in comparison to the secondaries. This effect is illustrated by normalizing all the spectra to unity before the resonance, as in figures 3(a)-(c). Thus, the low off-resonant background for the Auger lines results in an enhanced asymmetry, which can also be seen in the XMCD spectra in figure 3(d).
In order to evaluate the use of Auger electrons in imaging magnetic domains, a spontaneously nanostructured Fe film on W(110) was obtained by annealing a uniform 12 ML thick film to above 600 K [21]. The XMCD-PEEM image obtained from such a surface by the secondary photoelectrons is shown in figure 4(a), along with the image obtained by the Auger electrons at 595 eV as seen in figure 4(b). Black and white regions reflect [110] magnetic domains aligned parallel and antiparallel to the x-ray beam propagation, respectively. Note that the Auger-XMCD image is the difference between two images with opposite helicities without normalizing the result to their sum. The reason is that the low-intensity level in parts of the images results in additional noise upon division by the sum image. Nevertheless, low background level, insensitivity to work function changes and the long inelastic electron mean free path render imaging using Auger electrons an attractive alternative.
On the other hand, the large circular dichroism in x-ray absorption is accompanied by more information due to the fine details in the dichroic response of different Auger features. The large XMCD signal due to absorption can be appreciated in the Auger spectrum in the inset of figure 5(a), in which the entire spectrum changes in intensity by about 30% by switching the helicity of circular polarization. However, by normalizing out this overall intensity change, one can notice a difference between different Auger components as seen in figure 5(a). In particular, the largest asymmetry is at kinetic energies near the 3 P peak, with a long tail extending towards low kinetic energy. This signal is highly similar to the dichroism in LMM Auger emission even when the magnetization direction is perpendicular to the beam propagation direction, mostly based on the exchange coupling between the 3p holes and the spin-polarized valence band [22].

Conclusions
With the advent of synchrotron sources, the capabilities of AES, which is a well-established tool for surface chemical analysis, have expanded to include the exploration of electronic effects and femtosecond electron dynamics using corehole lifetime as an internal clock in resonant AES. Our study applied resonant Auger spectromicroscopy to ultrathin Fe films on W(110), tracking the transition from Raman to classical Auger behavior as a function of Fe thickness. Raman-Auger transition was found to be much sharper for monolayer and bilayer Fe as compared to thicker films, which showed a more gradual transition from Raman to classical Auger behavior. Furthermore, we conducted comparative XMCD spectromicroscopy studies on ultrathin Fe films using both secondary photoelectrons and Auger electrons. In particular, XMCD-PEEM images acquired with Auger electrons showed discreet Auger emission spectra at the Fe L 3 resonance (hν = 708.1 eV) measured with right and left circularly polarized x-rays, shown after normalizing to the low kinetic energy intensity. Inset shows the same spectra before normalization, which differ considerably in intensity due to the XMCD effect, shown as the area highlighted in pink. (b) Asymmetry between the two Auger spectra after normalization.
image quality and the XMCD spectra displayed a lower background in comparison to those acquired using secondary electrons. Thus, Auger spectromicroscopy holds promise for characterizing ultrathin films and studying electronic dynamics at the nanoscale.

Data availability statement
The data cannot be made publicly available upon publication because no suitable repository exists for hosting data in this field of study. The data that support the findings of this study are available upon reasonable request from the authors.