To submit to PRB (2013) Chemical local order signature on magnetism in CoPt and FePt nanoalloys A. Hillion1†, G. Khadra1, A. Cavallin2, A. Tamion1, J. Tuaillon-Combes1, F. Tournus1, E. Otero3, P. Ohresser3, S. Rusponi2, H. Brune2, A. Rogalev4, F. Wilhelm4 and V. Dupuis1*, 1 ILM, UMR 5306 CNRS/Université de Lyon, F-69622 Villeurbanne cedex Institute of Condensed Matter Physics, EPFL, 1015 Lausanne, Switzerland 3 Synchrotron SOLEIL, L’Orme des Merisiers, Saint-Aubin BP 48, F-91192 Gif-sur-Yvette Cedex 4 European Synchrotron Radiation Facility, BP 220, F-38043 Grenoble Cedex 2 Abstract. By combining the high photon flux and the chemical selectivity of several synchrotron radiation facilities, X-ray absorption spectroscopy has been intensively used to study finite size effect on the local structure and magnetism of CoPt and FePt clusters upon annealing-driven transition to chemically ordered L10like phase. Reversely to the bulk alloys upon chemical ordering, X- ray magnetic circular dichroism (XMCD) investigations at each L2, 3 edge, revealed a significant increase of both spin and orbital of Co (resp. Fe) and Pt magnetic moments after thermal ordering transition. Moreover these values largely overpass all the reported values known in the literature. Nevertheless, limited Magnetic Anisotropy Energy (MAE) enhancements have been obtained from SQUID magnetometry measurements on the same chemically ordered CoPt and FePt nanomagnets samples. In view to try to relate such magnetic behaviour to local structures in nanoalloys, we performed extended x-ray absorption fine structure (EXAFS) experiments and simulations at Co, Fe and Pt edges. In agreement with previous Vienna ab-initio simulation package (VASP) calculations, we experimentally evidenced a systematic element-specific dependence of the local atomic relaxations in asprepared and annealed CoPt and FePt clusters. We claim that this strong distortion in pure magnetic planes which do not match the underlying Pt layer, could explain the deviation to bulk alloys magnetic behaviours. * Electronic address : Veronique.Dupuis@univ-lyon1.fr 1 Introduction Magnetic nanoalloys attract a lot of attention because they offer the possibility to tune the magnetic moments and the magnetic anisotropy energy (MAE) probably up to the ultimate density storage limit. In particular, an extremely high magnetocrystalline anisotropy is expected from the stacking of pure Co (or Fe) and Pt atomic planes in the (001) direction for CoPt (or FePt) bulk alloys in the chemically ordered L10 phase. Nevertheless, even if a well chemical order can be observed at nanosize, the achievement of a consequent MAE enhancement without any coalescence after annealing-driven transition, remains so far absent [1, 2, 3, 4]. The atomic structure and magnetic moment of CoPt and FePt nanoparticles have been experimentally observed to differ from the corresponding bulk materials in a favourable way or not, due to small size effects as peculiar symmetry [5], partial chemical ordering [6], surface segregation [7],.. . Numerous theoretical works have been performed to try to explain exotic structure [8, 9, 10, 11] or electronic properties [12,13] observed in nanoalloys by integrating a great number of parameters. Recently, we evidenced element-specifc relaxation in size-selected CoPt clusters with 2-4 nm diameter range, from experiment to theory confrontation [14] . In this paper, in one hand, by using powerful X-ray magnetic circular dichroism (XMCD) investigations on FePt and CoPt clusters embedded in amorphous carbon matrix upon annealing, we found a significant increase of Fe, Co and Pt magnetic moments for the chemically ordered L10-like clusters structures, reversely to the bulk values. In the other hand, the magnetic anisotropy energy (MAE) of annealed samples determined from SQUID magnetometry is only twice that of the as-prepared samples, i.e. one order of magnitude smaller to what is expected for the L10 bulk phase. So we decided to quantify the Co-Co (resp. Fe-Fe), Co-Pt (resp. Fe-Pt) and Pt-Pt nearest neighbour interatomic distances from EXAFS simulations and compared to theoretical calculations obtained from Vienna ab-initio simulation package (VASP) in the same size range. From complementary x-ray absorption spectroscopy experiments at each specific Co (resp. Fe) and Pt environment, we have been able to relate structural and magnetic local order in bi-metallic CoPt (resp. FePt) nanoclusters in the 3 nm range. 2 Experimental procedures 1) Sample Synthesis and Transmission electron Microscopy characterization Bi-metallic clusters are preformed in the gas phase thanks to a laser vaporization source working in the Low Energy Clusters Beam Deposition (LECBD) regime. Briefly, a YAG laser ( = 532 nm, pulse duration 8 ns, frequency 30 Hz) is used to vaporize a mixed equiatomic (CoPt or FePt) target rod. Simoustaneously, a continuous flow of inert gas (He, 30 mBar), is injected to rapidly cool the generated plasma and to nucleate clusters submitted to a supersonic expansion under vacuum. Moreover our apparatus is equipped with a quadrupolar electrostatic mass-deviator allowing us when necessary to deposit mass-selected clusters in an UHV deposition chamber [5, 6] . The matrix is evaporated with an electron gun working under UHV conditions (base pressure of 5.10-10 Torr). Clusters and atomic matrix beams are simultaneously codeposited in UHV on 45°-tilted substrate front of both independent arriving beams. We have focused our attention on CoPt and FePt nanoparticles embedded in an inert carbon matrix, in order to preserve and investigate the intrinsic cluster surface properties. Using LECBD, we have been able to synthesize mixed FePt nanoparticles assemblies with a log normal size distribution centered on 3 nm in diameter which cover the range of mass-selected CoPt clusters present in this paper. The nanoparticles are 1%-diluted in volume to quite avoid magnetic interaction among nanoparticles. The idea was to study structure and magnetism of both FePt and CoPt nanoparticles (NPs) embedded in amorphous carbon matrix before and after an optimum annealing of 2 hours at 750 K under vacuum without any coalescence. First of all, structural characterizations have been performed using Transmission Electron Microscopy (TEM) on isolated CoPt and FePt clusters deposited on an amorphous carbon coated grid (then protected by a carbon thin film), before and after annealing under High Vacuum. The clusters' size distributions are experimentally determined from transmission electron microscopy (TEM) observations. For FePt samples prepared without mass selection, the size-histogram has been fitted with a log normal size distribution centered on ~ 2.6 nm in diameter and relative size dispersion around 40% [1]. While for mass-selected CoPt sample obtained thanks to the electrostatic quadrupole deflector, we used a Gaussian shape centered on 3nm with a size-dispersion lower than about 8% [15]. As upon 2 hours annealing under vacuum at 750 K, the size histogram fitted with a similar Gaussian curve, we claimed that there is no coalescence [15]. High Resolution TEM-observations and simulations on both CoPt and Fept NPs revealed that upon annealing a transition occurred from as-prepared A1 fcc-structure to a tetragonal chemically ordered L10 phase with a quasi perfect order parameter [15]. Notice that a perfect fcctruncated CoPt or FePt octahedrons predicted from the Wulff theory in our range of interest with a diameter of Dm = 1.8 nm, (resp. Dm = 2.7 nm) contains 201 atoms (resp. 586 atoms) and a number of atoms in the first surface-monolayer equal to 60.7 % (resp. 46.4 %). (a) (c) (b) (d) Fig. 1. HRTEM images of FePt cluster as-prepared (a) and annealed (c) FePt clusters coated wirh thin carbon film. In the fast Fourier Transform (b) and (c ), we clearly evidenced peaks and angles corresponding respectively to A1 fcc and L10 FePt phase transition upon annealing without coalescence. Representative HRTEM micrographs are presented in Fig.1 even if a few percent of decahedron and multitwined shape has also been detected [5]. 2) Magnetic Results – XMCD experiments Then, x ray magnetic circular dichroism (XMCD) signal has been measured on as-prepared and annealed bimetallic nanomagnets CoPt and FePt samples, respectively at the Co L-edge on SOLEIL – DEIMOS and at the Fe -Ledge on X’Treme – SLS beamline (see resp. Figure 2 and 3). Moreover, taking into account that induced magnetic moments resulting from 3d-5d proximity effects are also expected in non magnetic metal, we performed (XMCD) measurements at Pt-L edge on ID12 - ESRF beam line on the previous samples (see resp. Figure 4 and 5). In table I, the mean orbital and spin magnetic moments per Co, Fe and Pt atom, L and S have been determined using the wellknown sum rules16, 17 and the numbers of hole per Co, Fe and Pt atoms estimated from theoretical band structure calculations for L10 CoPt and FePt by applying the sum rules to theoretical XMCD spectra (n h Co = 2.628, n h Fe = 3.705 and n h Pt = 2.369) 18. TABLE I. Magnetic size diameter (dm and ), MAE (Keff) and maximum susceptibility (Tmax) determined from SQUID measurements. Co, (resp. Fe) and Pt spin and orbital magnetic moments from experimental XMCD measurements where the values L and S were calculated using the numbers of hole per Co, Fe and Pt atoms estimated from theoretical band structure calculations for L10 CoPt and FePt by applying the sum rules to XMCD spectra (n h Co = 2.628, n h Fe = 3.705 and n h Pt = 2.369). Average interatomic distances (R P.theoretical Ohresser in Å) and obtained on as-prepared) and annealed CoPt (resp. FePt) clusters (3nm) from EXAFS simulations at both Co (resp. Fe)-K and Pt-L edges. The site specific L10 lattice parameter ratio [c/a] is given for the annealed samples. Notice that the best fit has been obtained for relaxed geometry chemically ordered CoPt and FePt NPs with 201 and 586 atoms respectively from VASP calculations by assuming truncated octahedral structures in agreement with TEM observations. 3 SQUID dm (nm) EXAFS Co-(resp. Fe) K edge R (Å) EXAFS Pt-L edge XMCD Pt-L2. 3 SB/at.) LB/at.) L/S SQUID Keff (KJ/m ) K SQUI D Tmax (K) Å XMCD Co- (resp. Fe) L2. 3 SB/at.) LB/at.) L/S [c/a] [c/a] Å R (Å) CoPt asprepared TM 3.-- nm ? Co-Co: 2.52 Co-Pt: 2.58 0.012 Pt-Pt: 2.70 Co-Pt: 2.57 0.0045 1.67 0.13 0.077 0.47 0.07 0.150 218 37% 15 ? CoPt annealed TM 3.-- nm ? Co-Co: 2.57 Co-Pt: 2.62 Pt-Pt: 2.71 Co-Pt: 2.63 22 ? [0.92] 0.52 0.10 0.192 293 28% [1.03] 1.98 0.20 0.101 0.009 0.004 Fe-Fe: 2.57 Fe-Pt: 2.63 0.012 Pt-Pt: 2.71 Fe-Pt: 2.63 0.008 1.71 0.11 0.06 ? 15 ? _ Fe-Fe: 2.63 Fe-Pt: 2.63 Pt-Pt: 2.70 Fe-Pt: 2.64 ? [0.93] 0.57 0.07 0.13 ? [1.00] 2.58 0.18 0.07 0.012 0.008 FePt asprepared TM 3 nm FePt annealed NT 3 nm 40 % ? ? a) a) b) b) Fig. 2. Comparison between the experimental XMCD signal at the Co-L edge for the mass-selected CoPt sample before (a) and after annealing (b). Fig. 3. Comparison between the experimental XMCD signal at the Fe-L edge for the FePt sample before (a) and after annealing (b). Fig. 4. Comparison between the experimental XMCD signal at the Pt-L edge for the mass-selected CoPt sample before (a) and after annealing (b). Note that, even if the samples are made of randomly oriented crystallized nanoparticles, the magnetic dipole term T, which reflects the asphericity of the spin moment distribution around the absorbing atom, averages to zero only for 3d metal sites19. Thus, in table 1, S represents an effective spin moment at the Pt site (where T, can be significant) while is the true spin magnetic moment at the Co and Fe edges. We first verified that the mass-selected CoPt sample at the Co-site, present an analogous evolution compared to the previous samples of ref 1 (see figure 2). In the margin of error, estimations of the spin (1.7/2.0 B/at.) and orbital (0.13/ 0.20 B/at.) magnetic moment for the (as prepared/annealed) CoPt sample, lead to the same magnetic moment increase upon annealing. While the chemically disordered to the L10 phase has been clearly evidence in CoPt nanoparticles, this anomalous net Co magnetic moment enhancement is in contradiction with the bulk behaviour where the magnetization is decreasing when going from A1 to L10 phase20. To verify that this effect could not be due to an extensive Pt magnetic moment reduction which could compensate the Co magnetic moment variation, XMCD signal has been measured on the same sample at the Pt sites (see Figure 4). In this case, induced magnetic moments originate from the hybridization of the Pt 5d orbitals with the Co spin-polarized 3d states, we estimated the effective spin (0.47/0.52 B/at.) and orbital (0.07/ 0.10 B/at.) induced magnetic moment for the same (as prepared /annealed) CoPt sample. Once more, there is a striking change in the induced moment upon annealing: the S value is enhanced by 11%. This effect definitively invalidates the previous assumption on the Co magnetic moment compensation. Moreover, it is worth noting that spin momentum at both Co and Pt sites in annealed CoPt clusters (resp. equal to 2.00 and 0.52 B/at.) are larger than the calculated values for L10 CoPt bulk phase (1.85 and 0.37 B/at.)18. In addition, due to the fairly large spinorbit coupling of the Pt 5d electrons, the orbital moment largely contributes to the total moment on the Pt site and increase significantly as L/S ratio at Pt edge upon Fig. 5. Experimental XMCD signal at the Pt-L edge for the FePt sample after annealing. annealing. Because of the largely directional chemical bonding of Co atoms to Pt, anisotropic chemical local order, L10 like phase, plays a key role in enhancing the volume anisotropy as compared to the isotropic chemically disordered FCC structure. Then, we also performed XMCD measurements on FePt samples (see Figure 3 and 5). In figure 3a, one can notice a small multiplet in the Fe-L3 absorption edge for the asprepared FePt sample, which transforms into a small shoulder upon annealing (see Fig; 3b). It is due to the different chemical environment of Fe atoms at the NP-C interface as we will see in details later. By regarding the evolution of the mean spin magnetic moments per Fe atom, reported in table I, one can notice that the initial spin iron moment is smaller than in pure Fe-bulk site (2.12B/at.) but larger at both Fe and Pt sites in annealed FePt clusters (resp. equal to 2.58 and 0.57 B/at.) than the calculated values for L10 FePt bulk phase (2.50 and 0.47 B/at.)18. This must be undoubtedly related to finite-size effect in nanoalloys. According to Stern and Gerlach experiments on pure free Co and Fe clusters, the deviation from the bulk behaviour occurs around 400 atoms21. In our case, the mean CoPt and FePt cluster size corresponds to 200-600 atoms. So we can claim that finite size effects in deposited bimetallic clusters have been clearly evidenced for the first time without any in-situ removing native oxide shell7. In addition, the carbon matrix provides a efficient external degradation and oxidation protection of nanoparticles, in particular one of the best thermal resistant due to a preferential carbon graphitization upon high temperature annealing under vacuum 6, 22, 23 . SQUID measurements Finally, the magnetic behavior of the cluster-assembled films has also been studied using SQUID magnetometry (Superconducting Quantum Interference Device). An accurate analysis has been reached using our recently developed “triple fit” method where the ZFC/FC curves and a room temperature magnetization loop are simultaneously adjusted using a semi-analytical model [24] . Our method is based on an improved description of the susceptibility curves with a progressive crossover between the blocked regime and the superparamagnetic regime instead of the usual abrupt transition model, and takes into account the experimental temperature sweeping rate. By fitting simultaneously the entire magnetic curves, we can reach a reliable and accurate determination of the most important characteristics of a sample, namely the particle magnetic size distribution and MAE [10]. In Fig. 6, we present these 3 experimental SQUID curves and their fits obtained on both as-prepared and annealed CoPt samples. Notice that in the superparamagnetic regime, the magnetization curves at 300 K are found identical for both samples. Fig. 6: Experimental and simulated magnetic susceptibility curves from Field Cooled, Zero-Field-Cooled protocole (top) and hysteresis loop at 300 K (bottom) of as-prepared and 750 K annealed CoPt nanoparticles embedded in carbon matrix. The effective anisotropy constant and the clusters magnetic size distribution are obtained by a simultaneous triple fit procedure of the FC, ZFC and a room temperature magnetization loop [10]. From systematic “triple fit” simulations [10], we have been able to deduce the average magnetic diameter (Dm) taking into account the TEM distribution, the mean anisotropy constant (Keff) and their corresponding standard deviation and(see TAB. I). One can note that the median magnetic diameter Dm, is equal to the observed TEM one and remains constant upon annealing. This means that there is no magnetically dead interface between the cluster surface and the matrix and that the cluster size is conserved at least up to 750 K. The main difference comes from the MAE evolution between the as-prepared and annealed CoPt sample where the median effective anisotropy constant (Keff) increased and its corresponding dispersion slightly decreased. This last dispersion decrease is well explained by the appearance of the chemical order at nanosize [ 15]. But, the MAE value is one order of magnitude smaller that what is expected for the L10 CoPt bulk. As-prepared FePt sample in carbon matrix has the same Tmax as CoPt… A suivre 3) Structural characterization Then, we have studied the local environment in these magnetic nanoalloys on the CRG-FAME and ID12 beamlines of the European Synchrotron Radiation Facilities at Grenoble. Extended x-ray absorption fine structure (EXAFS) at both Co (resp. Fe)- K and Pt-L edge has been performed in the fluorescence mode on asprepared and annealed on CoPt (resp. FePt) samples. From FEFF Fit simulations, we clearly show that the CoCo/Co-Pt (resp. Fe-Fe/Fe-Pt) coordination ratio changes upon annealing from 1 to 2 as expected for the A1 to L1 0 transition with contracted lattice parameters related to the bulk phase. After annealing, EXAFS experiments performed at both Co (resp. Fe) K and Pt L edges, on the same annealed CoPt (resp. FePt) sample are presented in figures 7 and 8. The best simulation has been obtained for a perfect chemically ordered CoPt (resp. fePt) packing (L10 phase) with a unique Co-Pt (resp. Fe-Pt) first neighbour distance at both edges. But surprisingly, we found a reverse apparent c/a ratio at both edges, namely c/a = 1.03 (resp. c/a = 1) at the Co-K (resp. Fe-K) edge and c/a = 0.92 (resp. 0.93) at the Pt-L edge for annealed CoPt sample (resp. FePt sample) see Table I. From previous performed spin-polarized densityfunctionnal calculations using the Vienna ab initio simulation package (VASP) with first principles magnetic and structural optimizations [25], applied to perfect CoPt clusters in the same size range, we found 3 different Gaussian relaxed distance distributions for Co-Co, Co-Pt and Pt-Pt both in chemically disordered and L1 0 ordered phases. In particular, while in the bulk phases, an equal Co-Co (resp. Fe-Fe) and Pt-Pt Dirac distance distribution is expected, a strong Co-Co distance dispersion has been found for relaxed clusters even in the chemically ordered phase. The tendency of Co nearest neighbour bonds to be shorter than the Pt ones results in structural stress, which in a finite CoPt cluster can be more easily distorted by moving Co atoms (than Pt atoms) in the surface faces. As seen in Table I, one can notice the very good agreement between EXAFS simulation and VASP predictions. As expected from the respective TEM sizedistribution, the FePt samples presents a larger DebyeWaller value compared to the size selected CoPt sample. An other important difference existing in the FePt samples compared to CoPt, is the fact that it is necessary to consider C atoms neighboring in the Fe-environment, especially in as-prepared FePt phase. The contribution of interstitials carbon chemical environment at the FePt clusters surface is noticeable by the smoothing of the EXAFS oscillations compared to the CoPt curves (see Figures 7 and 8). In fact, the enthalpy of carbide formation is close to zero in Fe fcc and increases by Fig.7. Comparison between the experimental EXAFS signal (dots, contribution of the NN peak only) and simulated curves (solid lines) with c/a = 1.03 at the Co-K edge (a) and with c/a = 0.92 at the Pt-L edge (b) obtained on L10 3 nm in diameter CoPt clusters in very good agreement with VASP calculations. reducing the particle size26, 27. This metastable carbide is probably at the origin of the magnetically dead interface layer found in the as-prepared FePt sample from XMCD at Fe edge. Annealing the samples at 750 K favors the graphitization of the matrix and removes the dead layer without deteriorating the nanoparticle size distribution 23. Moreover, we have to notice that this demixing has not a damage effect on the magnetic moment of annealed sample. Discussions concerning MAE and magnetism from SQUID magnetometry measurements on FePt sample, are in progress … 3 Conclusion The VASP ab-initio calculations fully confirm the experimental structural trends in the size-range of our nanoalloys, providing a detailed account of the element specific local relaxations which bring back together the EXAFS experimental results at both element edges. Indeed, this element-specific dependence of the local atomic relaxations in CoPt and FePt clusters lead to a strong distortion in pure magnetic planes. The fact that Fig. 8. Comparison between the experimental EXAFS signal (dots, contribution of the NN peak only) and simulated curves (solid lines) with c/a = 1.00 at the Fe-K edge (a) and with c/a = 0.95 at the Pt-L edge (b) obtained on L10 FePt clusters in very good agreement with VASP calculations. pure Co and Fe layers do not match the underlying Pt layer is an specifity of finite size effects in nanaoalloy, certainly relate to the low MAE results obtained on chemically ordered L10-like nanoalloys. We claim that we obtain new insights on the correlation between magnetic properties and short- or long-range chemical order in nanoalloys. We clearly demonstrate that the magnetic moment increase while the MAE for nearly perfect chemical ordered nanoparticles assemblies is one order lower to what is expected for the bulk. In a previous paper [28], we have shown that the MAE dispersion in pure Co clusters essentially comes from the effect of additional facets and is relatively small. But in nanoalloys, the size and shape dispersion is not the unique source of MAE dispersion. Indeed, since the anisotropy enhancement in as-prepared CoPt compared to pure Co clusters is due to the presence of Pt atoms, the dispersion of the magnetocrystalline anisotropy (which depends on the neighborhood of each Co atom) increases with the number of possible chemical arrangements. It is the reason why the MAE of chemically disordered CoPt particles is quite large even if mass-selected clusters have a small size dispersion (8%) and a highly symmetrical shape (regular truncated octahedron) [15]. As long as a well-defined and high enough degree of chemical order can be reached, the multiplicity of atomic configurations is strongly reduced and the MAE dispersion decreases while its median value increases29. Discussions concerning MAE and magnetism in FePt sample, are in progress … To go further, non-collinear calculations, including the spin-orbit coupling in order to quantitatively access to the MAE values are in progress for such relaxed nanoalloys (Ab-initio+ tight-binding calculations). Acknowledgment The authors are grateful to A. Ramos, H. Tolentino, M. de Santis and Olivier Proux for stimulating discussions and for their help during experiments on the French CRG-D2AM and BM30b-FAME beamlines at ESRF and to J. Dreiser, C. Piamonteze and F. Nolting from Swiss Light Source for their investment on the X’Treme beamline. Support is acknowledged from both GDR CNRS 3182 and COST-STSM-MP0903-7318 on Nanoalloys † Current address: Laboratory of Solid State Physics and Magnetism, K.U. Leuven, Belgium References 1 F. Tournus, A. Tamion, N. Blanc, A. Hannour, L. Bardotti, B. Prevel, P. Ohresser, E. Bonet, T. Epicier, and V. Dupuis, Phys. Rev. B 77, 144411 (2008). 2 F. Tournus, N. Blanc, A. Tamion, M. Hillenkamp and V. Dupuis, Phys. Rev. B 81, 220405(R) (2010). 3 C.-b. Rong, N. Poudyal, G. S. Chaubey, V. Nandwana, R. Skomski, Y. Q. Wu, M. J. Kramer, J. P. Liu, J. Appl. Phys. 102, 043913 (2007). 4 X. Sun, Z. Y. Jia, Y. H. Huang, J. W. Harrell, D. E. Nikles, K. Sun, and L. M. Wang, J. Appl. Phys. 95, 6747 (2004). 5 F. Tournus, K. Sato, T. J. Konno, T. Epicier, V. Dupuis, Phys. Rev. Letters 110, 055501 (2013) 6 N. Jaouen, D. Babonneau, J. M. Tonnerre, D. Carbone, F. Wilhelm, A. Rogalev, T. K. Johal, and G. van der Laan, Phys. Rev. B 76, 104421 (2007). 7 Antoniak, J. Lindner, M. Spasova, D. Sudfeld, M. Acet, M. Farle, K. Fauth, U. Wiedwald, H.-G. Boyen, P. Ziemann, F. Wilhelm, A. Rogalev, and S. Sun, Phys. Rev. Lett. 97, 117201 (2006). 8 G. Rossi, R. Ferrando, and C. Mottet, Faraday Discuss. Chem. Soc. 138, 193 (2008). 9 M. E. Gruner, G. Rollmann, P. Entel, and M. Farle, Phys. Rev. Lett. 100, 087203 (2008). 10 P. Entel, M. E. Gruner, G. Rollmann, A. Hucht, S. Sahoo,, T. Zayak, H. C. Herper, and A. Dannenberg, Philos. Mag. 88, 2725 (2008). 11 P. Andreazza, C. Mottet, C. Andreazza-Vignolle, J. Penuelas, H. C. N. Tolentino, M. De Santis, R. Felici, and N. Bouet, Phys. Rev. B 82, 155453 (2010). 12 C. Barreteau and D. Spanjaard, J. Phys.: Condens. Matter 24 (2012) 406004 13 R. Cuadrado and R. W. Chantrell, Phys. Rev. B 86, 224415 (2012) 14 V. Dupuis, N. Blanc, L-E. Diaz-Sanchez, A. Hillion, A. Tamion, F. Tournus, G-M. Pastor, A. Rogalev and F. Wilhelm, Eur. Phys. J. D (2013) 67: 25 15 F. Tournus, N. Blanc, A. Tamion, M. Hillenkamp, V. Dupuis, J. Magn. Magn. Mater. 323, 1868 (2011) V. Dupuis, N. Blanc, F. Tournus, A. Tamion, J. TuaillonCombes, L. Bardotti, O. Boisron, IEEE Trans. Magn. 47, 3358 (2011) N. Blanc, F. Tournus, V. Dupuis, T. Epicier, Phys. Rev. B 83, 092403 (2011) as Brief Report 16 B. T. Thole, P. Carra, F. Sette, and G. van der Laan, Phys. Rev.Lett. 68, 1943 (1992). 17 P. Carra, B. T. Thole, M. Altarelli, and X. Wang, Phys. Rev. Lett. 70, 694 (1993). 18 I. Galanakis, M. Alouani and H. Dreysse: J. Magn. Magn. Mater. 242–245 (2002) 27 19 J. Stöhr, J. Electron Spectrosc. Relat. Phenom. 75, 253 (1995) 20 R. A. McCurrie and P. Gaunt, Philos. Mag. 13, 567 (1966) 21 I. M. L. Billas, A. Chatelain, and W. A. de Heer, J. Magn. Magn. Mater. 168, 64 (1997) 22 A. Hillion, A. Cavallin, S. Vlaic, A. Tamion, F. Tournus, G. Khadra, J. Dreiser, C. Piamonteze, F. Nolting, S. Rusponi, K. Sato, T. J. Konno, O. Proux, V. Dupuis, and H. Brune, Phys. Rev. Lett. 110, 087207 (2013) 23 A. Tamion, M. Hillenkamp, A. Hillion, F. Tournus, J. Tuaillon-Combes, O. Boisron, S. Zafeiratos and V. Dupuis, J. Appl. Phys. 110, 063904 (2011) 24 A. Tamion, M. Hillenkamp, F. Tournus, E. Bonet, V. Dupuis, Appl. Phys. Lett. 95 062503 (2009) 25 N. Blanc and al., accepted to Phys. Rev. B (2013) 26 A. Tolvanen, A. V. Krasheninnikov, A. Kuronen, and K. Nordlund, Phys. Status Solidi (c) 7, 1274 (2010) 27 A. R. Harutyunyan, N. Awasthi, A. Jiang, W. Setyawan, E. Mora, T. Tokune, K. Bolton, and S. Curtarolo, Phys. Rev. Lett. 100, 195502 (2008). 28 M. Jamet, W. Wernsdorfer, C. Thirion, D. Mailly, V. Dupuis, P. Mélinon, and A. Pérez, Phys. Rev. Lett. 86, 4676 (2001) 29 F. Tournus, S. Rohart and V. Dupuis, IEEE Transactions on Magnetics 44 3201 (2008)