The beauty of the XMCD technique is the element specificity. That is, one may probe separately contributions from various magnetic elements in a single sample. As an example, we plot in Fig. 1 the absorption coefficient µ(E) as a function of the energy of the incident x-ray synchrotron beam for one Co/Cu/Ni/Cu(001) trilayer (see inset). One encounters successively the strong absorption resonances of the incident beam at the L_3,2 edges of Co and Ni. By using circularly polarized x-rays one may record the difference in the absorption spectra between left and right circularly polarized x-rays. This difference, the so-called XMCD signal is a measure of the magnetization. After proper analysis by means of the sum rules the magnetic moments of Co and Ni and a separation into spin and orbital moments is possible, see e.g. [Refs.166,203] . In this process the polarization degree of the x-rays and the number n _hof the d-holes for Co and Ni are necessary inputs. In our measurements we use the electron yield detection mode and the fluorescence detection mode. Spectra from single layers and trilayers were recorded at BESSY, the synchrotron facility in Berlin, Germany and from multilayers at the ESRF in Grenoble, France.

• Effects of interlayer exchange coupling on the element-specific magnetizations in Co/Cu/Ni trilayers

Magnetic layers interact through non-ferromagnetic spacers via the interlayer exchange coupling J_inter (Fig. 2). In the end of the 80's and the 90's a lot of research effort was put on trilayers and multilayers due to the novel properties which were related to J_inter, such as the Giant Magnetoresistance.  The main activities were focusing on the optimization of the properties useful for technological applications at room temperature. Recent element-specific temperature-dependent measurements on Co/Cu/Ni trilayers as prototype for multilayers elucidate the behavior of fundamental observables such as the sublayer magnetizations under the influence of J_inter. In Fig.3(a) we see the XMCD spectra at the L_3,2 edges of Ni and Co in a Co/Cu/Ni/Cu(001) trilayer at 290 and 336 K [Refs. 203,227] . Co and Ni XMCD spectra at 290 K reveal ferromagnetism in both layers (dotted lines). However, at 336 K Co still has non-zero XMCD signals, while, within our experimental resolution, Ni shows no ferromagnetic response (solid lines).This means that Ni has entered the paramagnetic phase. In Fig. 3(b) the temperature-dependence of the remanent magnetizations of both Ni and Co are presented near the temperatures of interest. It is clearly shown that Ni shows no ferromagnetism above T ^*_C(Ni)=308 K. The T_C of Co was found to be about 340 K. These measurements show that, despite the very thin Cu spacers allowing for a strong interlayer exchange coupling between Co and Ni, the two layers enter the paramagnetic phase at different temperatures. The XMCD technique may also probe the sign of J_inter. Note in Fig. 4 the difference of the sign ofNi spectra in trilayers with Cu spacers of different thickness. From these spectra J _inter is determined to have the opposite sign in the two trilayers [Ref. 212] . The observation of antiferromagnetic (AFM) coupling in trilayers with such thin Cu spacers guarantees that the physics is dominated by the interlayer coupling and not from pinhole problems.
 
 

Fig. 3

Fig. 4

Figure 5 depicts element-specific remanent-magnetization curves for Ni and Coin bi- and trilayers. The Ni magnetization in the bilayer (open circles) vanishes at a temperature of about 270 K which corresponds to the temperature T _C^Ni, since Ni is in-plane magnetized. After thedeposition of Co on the top of the Cu/Ni bilayer the temperature T^*_C ^Ni where the Ni magnetization (closed circles) vanishes is increased by DeltaT_Ni >>+40 K. The Co magnetization (closed diamonds), on the other hand, vanishes at much higher temperatures [Refs. 206,212,252] . The shift of the orderingtemperature of Ni has been understood by considering the reduced dimensionality of the Ni layers as it is explained in [Ref. 229] . Taking into account the oscillatory nature of J_inter with the spacer thickness d, one would expect a similar oscillation of the increase DeltaT_Niwith d. Such a behavior is shown in Fig. 6 [Refs. 212,252] . The sign of DeltaT_Ni is always positive, implying that the ordering temperature of Ni is always increasing, independently of the FM or AFM coupling. The solidcurve in Fig. 6 is calculated with the theoretical parameters describing the two periods of oscillation of J_inter in an RKKY-like model and their experimental amplitude. Note the excellent agreement between the experimental data points and the theoretical curve without any adjustable parameter.
 

Fig. 5

Fig. 6

In Fig. 7 we present results from a trilayer where Co has lower Curietemperature than the Ni one. In Fig 7(a) we see the XMCD spectra at T=45K. Of particular interest is the large increase in the Ni XMCD before and after the evaporation of Co on the top. A clear picture of the magnetic behavior can be obtained from Fig. 7(b). First of all we see that the Co magnetization is larger than the Ni one at low temperatures and due to the low ordering temperature of Co the temperature-dependent magnetizations of Ni and Co cross each other just below 80 K. The peculiar shape of the Ni magnetization can be understood by the change of its easy axis due to the interlayer exchange coupling [Ref. 248] . This experiment demonstratesthat the coupling can be used for the fabrication of temperature-dependent magnetic switches, see also [Ref. 204] .

  Fig. 7

• Co/Cu/Ni trilayers near their Curie temperature in a small magnetic field 


The early experiments on Co/Cu/Ni trilayers were carried out in remanence. Effects that possibly arise due to the formation of magnetic domains near the ordering temperature, cannot be entirely excluded. Such effects have been observed for Ni films on Cu(100) [187]. Therefore, improved measurements in a small applied magnetic field were carried out. Figure 8 (left) shows the temperature-dependent magnetization of Ni in a coupled trilayer measured at H = 40 Oe both before (open circles) and after evaporation of the Co layer (solid circles). Here, a shift of the Ni ordering temperature of ca. 80 K is observed. This shift confirms the results of Refs. [206, 245], so that they cannot be addressed to effects of domain formation. The exchange coupling of the Co film acts on the Ni film not only as a static exchange field but it influences also by higher order contributions the dynamics of the fluctuating spins. Therefore, theory expects a tail in the temperature-dependent magnetization near the ordering temperature [229]. However, this tail was not observed in the earlier experiments [252]. Taking advantage of the high photon-flux together with the high circular polarization-degree at the new undulator beamlines at third generation synchrotrons small XMCD signals near the Curie temperature can be determined more precisely. Better statistics can be obtained in shorter time and therefore, measurements close to T_C^*Ni with a relativly high density of data points have become achievable. Figure 8 (right) shows the individual Ni and Co magnetization curves for a trilayer consisting of 4 ML Ni, 3.3 ML Cu and 2.3 ML Co. Both curves were measured under a small applied field. In these new experiments in a temperature range close to T_C, we observe a tail of the Ni magnetization which is clearly out of the noise. The results are in agreement with theoretical predictions. They demonstrate that spin fluctuations, i.e. higher order contributions
in the spin dynamics, are of more importance in ultrathin films than in 3D bulk magnetism.

  Fig. 8 

 
• Complete magnetic moment profile and magnetic anisotropy in Ni/Pt multilayers

The magnetic properties of multilayers depend essentially on the interface magnetic moments of the ferromagnetic and the non-ferromagnetic, but magnetically polarizable, constituents. Progress in the XMCD technique within the third generation synchrotron radiation facilities allowed us to construct a monolayer-resolved magnetic moment profile for Ni/Pt multilayers. The samples were prepared in the group of Prof. Flevaris at the Aristotle University of Thessaloniki,Greece. The XMCD measurements were performed at the ID12B (Ni L_3,2edges) and the ID12A (Pt L_3,2 edges) in ESRF by using the electron yield or fluorescence detection modes respectively. In Fig. 9(a) we see x-ray absorption (red) and XMCD spectra (green color) from a Ni₂/Pt₂ multilayer at both Ni and Pt edges. The indices are number of monolayers in each multilayer unit cell. A schematic representation for such a multilayer is shown in Fig. 9(b). By combining measurements for a series of Ni/Pt multilayers as it is explained in [Ref. 233] we are able to produce a complete monolayer-resolved magnetic moment profile for a Ni₆/Pt₅ sample (top panel of Fig. 9(c)). The result is compared to ab initio calculations (bottom panel) by D. Benea and H. Ebert . We see the same trends for the magnetic profiles deduced by experiment and theory. The discrepancies for the interface moments may be bridged up by considering a small amount ofintermixing (up to 30%) strictly placed at the interface [Ref. 241] . The magnetic moment profile is presented for the first time for both constituents of a multilayer in one experiment and in theory and it gives a new insight into the fundamental understanding of interface magnetism. Orbital magnetism is an important chapter of the ultrathin film magnetism. Few techniques may provide spin and orbital magnetic moments, these are the Ferromagnetic Resonance and the XMCD. The anisotropy of the orbital moment is directly linked to the magnetic anisotropy energy (MAE). In Fig. 10 we plot the angular dependence of the orbital moment of Ni in a Ni₂/Pt ₂ multilayer. While the anisotropic moment itself is small (0.008 µB/atom) the relative change between the easy and the hard axis is larger than 20% driving the total moment out-of-the film plane as it is necessary for applications such as the magneto-optic recording [Ref. 230]
 
 

Fig. 9

Fig. 10


 
• The violation of the 3rd Hund's rule in Fe/W multilayers

The magnetic properties of layered ferromagnets may be radically different than in the bulk or in alloys due to different local geometry and electronic structure. Consequently, in many cases new phenomena may arise in magnetic multilayers. In Fig. 11 we plot XMCD spectra recorded at the W L_3,2 edges in Fe/W multilayers. The samples were prepared in the group of Prof. Flevaris at the Aristotle University of Thessaloniki, Greece . One can see the high quality of the XMCD spectra despite the fact that the difference between right and left circularly polarized light is small as the large magnification factor of 'x50' reveals. Secondly, by the sign of the XMCD integral (red dashed line) we conclude that the spin and orbital moments of W are parallel though W has a less-than-half filled 5d shell and the3rd Hund's rule predicts antiparallel configuration between spin and orbital moments (green dashed line). This spectacular violation of the 3rd Hund's rule which is for the first time experimentally addressed in a solid maybe understood by the model presented in the lower panel of Fig. 11. Namely, there is a strong interlayer spin-orbit interaction between Fe and W which drives the W orbital moment in an opposite direction than the one that the intraatomic W spin-orbit interaction favors.
  Fig. 11
 

• Sharp magnetic moment profiles in Fe/V/Fe(110) trilayers

The Fe/V superlattice is one of the systems where controversial results between experiment and theory were published for the Fe and V magnetic moments. The main question for V was whether there is a long-range order polarization at the interface (experiment) or this polarization is of short-range order(theory). New experiments with controllable preparation conditions resulting in ideal-like interfaces with minute amount of interface atomic exchange, supported by a STM study in Fig. 12, help us to elucidate the previous controversy Ref[265]. The schematic representation of our trilayers and normalized absorption and XMCD spectra at the V and L_2,3 edges are presented in Fig. 12. From the positive onset of the V-XMCD spectra one concludes that the Fe and V moments at the interface couple antiferromagnetically. Note that the V spectra are of much higher quality than any other reported. The rapidly decrease of the V-XMCD signal with the V thickness give a first support to a short-range V polarization. They may be used to construct monolayer-resolved magnetic moment profiles.These show unambiguously a short-range order profile in agreement, for the first time, to theoretical predictions Ref[265,273]. The V moments are given in a relative scale due to core hole correlation effects which, consequently, do not allow us to apply the sum rules.


  Fig. 12
  

Even though the XMCD spectra reveal a large variety of different spectral features, magneto-optical sum rules are commonly applied to determine the ground state spin and orbital moments. In fact the application of sum rules is rather limited in particular for the early 3d transition metals where the absorption spectra suffer from core hole correlation effects as reflected by a transfer of spectral weight in favor of the L₂ edge in Fig. 13. In order to study the influence of the core hole on the magnetic information, we have investigated both the XAS and XMCD spectra towards Ti in a systematic fashion Ref. [292]. While the early transition metals are either non-magnetic (Ti, V) or antiferromagnetic (Cr) in their bulk configurations and reveal no XMCD signal, they show a net magnetic moment in thin films when intercalated in ferromagnetic Fe in a Fe/TM/Fe(110) trilayer. The resulting XMCD spectra are nearly free of noise. Note that the XMCD to XAS peak height ratio at the L₃ edge is about 3%, 7% and 9% for Ti, V and Cr, respectively, due to the smaller spin-polarization of the d valence states with regard to the ferromagnets Fe, Co and Ni. Moreover, detailed fine structures are brought to light at and above the L_2,3 edges.

 Fig. 13

For the analysis of the XMCD fine structure and the determination of the related ground state magnetic moments of the early 3d elements, we applied an alternative concept Ref[269] which compares the experimental data with ab initio theory by J. Minar and H. Ebert. The calculations are based on a fully relativistic spin-polarized Korringa-Kohn-Rostoker (SPR-KKR) Green’s function method, which is appropriate for itinerant magnetic systems. This concept is demonstrated in Fig. 14 for V in a Fe/V trilayer and a FeV disordered alloy serving as an experimental standard. Core-hole correlation effects were not included accounting for the deviation of the experimental branching ratio (close to 1:1) of the V-XAS (see top panels). Looking at the XMCD spectra (bottom panels), we find that nearly the full absorption fine structure is reproduced by theory. While the theory explains clearly the feature (S) at the high-energy side as a van Hove singularity in the V band structure, the experiment could falsely assign the satellite to contamination with oxygen. Note that V absorption spectra of ex situ samples are often truncated just below the oxygen K edge (~532 eV). The second prominent feature (A) is missed in the calculations and therefore, stems most likely from a multiplet feature. Here, it turns out surprisingly that the 3d electrons in less-than-half filled transition metals appear considerably correlated giving rise for the atomic-like effects.

  Fig. 14

For the experimental standard the SPR-KKR calculations also provide the induced spin and orbital moments in Fig. 15 that are related to the XMCD spectra Ref[269]. These values are consistent with independent polarized neutron studies (PNS). The negative sign indicates that the induced moment in V is coupled antiparallel to that in Fe. While the sum rule analysis holds in principle for Fe, the application of particular the spin sum rule to V leads to an erroneous result by a factor 5. This indicates the breakdown of the spin sum rule caused by the core hole interaction which become competitive to the smaller 2p spin-orbit coupling in early 3d transition metals and therefore, intermixes the L_2,3 contributions in the absorption spectra.

   Fig. 15