Magnetic anisotropy of L1_0-ordered FePt thin films studied by Fe and Pt L_{2,3}-edges x-ray magnetic circular dichroism
K. Ikeda, T. Seki, G. Shibata, T. Kadono, K. Ishigami, Y. Takahashi, M. Horio, S. Sakamoto, Y. Nonaka, M. Sakamaki, K. Amemiya, N. Kawamura, M. Suzuki, K. Takanashi, A. Fujimori
MMagnetic anisotropy of L -ordered FePt thin films studied by Fe and Pt L , -edgesx-ray magnetic circular dichroism K. Ikeda, T. Seki, G. Shibata, T. Kadono, K. Ishigami, Y. Takahashi, M. Horio, S. Sakamoto, Y. Nonaka, M. Sakamaki, K. Amemiya, N. Kawamura, M. Suzuki, K. Takanashi, and A. Fujimori Department of Physics, University of Tokyo, Hongo 113-0033,Japan a)2)
Institute for Materials Research, Tohoku University, Sendai 980-8577,Japan Institute of Materials Structure Science, KEK, Tsukuba 305-0801,Japan JASRI, 1-1-1 Kouto, Sayo, Hyogo 679-5198, Japan (Dated: 27 June 2017)
The strong perpendicular magnetic anisotropy of L -ordered FePt has been the subject ofextensive studies for a long time. However, it is not known which element, Fe or Pt, mainlycontributes to the magnetic anisotropy energy (MAE). We have investigated the anisotropy ofthe orbital magnetic moments of Fe 3 d and Pt 5 d electrons in L -ordered FePt thin films byFe and Pt L , -edge x-ray magnetic circular dichroism (XMCD) measurements for sampleswith various degrees of long-range chemical order S . Fe L , -edge XMCD showed that theorbital magnetic moment was larger when the magnetic field was applied perpendicular tothe film than parallel to it, and that the anisotropy of the orbital magnetic moment increasedwith S . Pt L , -edge XMCD also showed that the orbital magnetic moment was smaller whenthe magnetic field was applied perpendicular to the film than parallel to it, opposite to the Fe L , -edge XMCD results although the anisotropy of the orbital magnetic moment increaseswith S like the Fe edge. These results are qualitatively consistent with the first-principlescalculation by Solovyev et al. [Phys. Rev. B , 13419 (1995).], which also predicts thedominant contributions of Pt 5 d to the magnetic anisotropy energy rather than Fe 3 d due tothe strong spin-orbit coupling and the small spin splitting of the Pt 5 d bands in L -orderedFePt.PACS numbers: 75.50.Bb, 75.30.Gw, 78.20.Ls. a) Electronic mail: [email protected] a r X i v : . [ c ond - m a t . m t r l - s c i ] J un n recent years, the amount of data we deal with has been increasing and the information tech-nology has to adapt to the era of big data. To this end, the enhancement of the recording densityin magnetic recording media has been highly desired. In the magnetic recording media like harddisk drives, larger data capacity can be achieved by using materials with perpendicular magnetizationand by decreasing the bit size, but it also leads to the reduction of thermal stability. To maintainthe thermal stability, it is important to employ magnetic materials with larger magnetic anisotropyenergy (MAE). L -ordered FePt is one of such materials that have the largest MAE with the perpen-dicular easy magnetization axis, and is a candidate to achieve higher density of magnetic recordingmedia. The microscopic origin of the strong perpendicular magnetic anisotropy of L -ordered FePt,however, has not been fully understood yet despite its importance and extensive studies.In order to reveal the origin of the MAE of L -ordered FePt, magnetism at the Fe sites and thePt sites need to be studied by an element specific technique. Especially, the orbital magnetic mo-ments of Fe and Pt have to be investigated separately since it has been well known that the orbitalmagnetic moment plays an important role in magnetic anisotropy . X-ray magnetic circular dichro-ism (XMCD) at the L , edges is an ideal tool to investigate the magnetic properties of d electronsin transition-metal alloys in an element specific way. In the present study, we have studied the spin( m spin ) and orbital magnetic moments ( m orb ) of Fe 3 d and Pt 5 d electrons and their anisotropy usingFe and Pt L -edge XMCD and discuss the microscopic origin of the MAE of L -ordered FePt thinfilms.FePt thin films were grown on MgO (100) substrates by the ultrahigh vacuum dc-sputtering methodwith Fe and Pt targets. The stacked structure of the sample was MgO substrate/Fe (1 nm)/Au (30nm)/FePt (20 nm)/Au (2 nm). The 2nm-thick top Au layer was deposited as a cap layer and the bot-tom Au layer of 30 nm thickness beneath the FePt layer was inserted as a buffer layer to reduce theeffect of the lattice mismatch between FePt and the MgO substrate. The deposition temperature ( T S )and the annealing temperature ( T A ) ranged from room temperature to 600 ◦ C . Table I summarizes therelationship between sample-preparation conditions, the degree of long-range chemical order S esti-mated using x-ray diffraction , and the MAE. Both S and MAE increase with T S and T A , consistentwith the previous study . The MAE of the sample with S = 0 shows a small negative value, which isdue to errors in the estimation of the MAE from the M-H curve (Fig. S.1).Fe L -edge XMCD measurements were performed at the undulator beamline BL-16A1 of PhotonFactory, High Energy Accelerator Research Organization (KEK-PF), and the Pt L -edge XMCD mea-surements at the undulator beamline BL39XU of SPring-8. All the measurements were performed atroom temperature. The magnitude of the magnetic field was 5 T at BL-16A1 and 7 T at BL39XU. Themagnetic field was applied parallel or anti-parallel to incident x rays. In order to measure the in-planemagnetic moments, the incident angle θ of x rays measured from the sample surface (as shown in2 ABLE I. Relationship between the sample-preparation conditions, the degree of long-range chemical order S , and the magnetic anisotropy energy (MAE) of samples studied in the present work. The MAE has beenestimated from the M-H curves (Fig. S1). The shape anisotropy energy has been subtracted from the MAE.
Depositiontemperature ( ◦ C ) Annealingtemperature ( ◦ C ) Degree of long-rangechemical order parameter S MAE( MJ / m )300 600 0.7 ± ± ± the inset of Fig. 1 and Fig. 3) was set to 30 ◦ and 4.4 ◦ , respectively, for the Fe L -edge and Pt L -edgeXMCD measurements. For the Fe L -edge XMCD measurements, four x-ray absorption spectroscopy(XAS) spectra were taken by inverting the external magnetic field and the helicity of x rays inde-pendently, and then were averaged. For the Pt L -edge XMCD measurements, the helicity switchingmode was used. The external magnetic field was inverted and then the obtained two XMCD spectrawere averaged. Fe L -edge XAS and XMCD were measured in the total electron-yield detection modeand Pt L -edge XAS and XMCD were measured in the partial fluorescence-yield detection mode withsilicon drift detector.Figure 1 shows the XAS and XMCD spectra at the Fe L , edge of L -FePt thin films ( S = 0,0.5, 0.7) under magnetic field µ H = 5 T applied parallel (in-plane) and perpendicular to the films(out-of-plane). Each XAS spectrum has been normalized to the height of the L peak. All the spectrashow line shapes similar to that of metallic Fe , and no oxidation features are observed. As shownin Fig. 1(c), the XMCD spectra are different between the two field directions, showing significantanisotropies: the Fe L -edge XMCD intensity becomes more anisotropic with increasing S while theFe L -edge intensity remains nearly isotropic. According to the XMCD sum rules , this behaviorindicates the increase of the anisotropy of the orbital magnetic moment with S . By applying theXMCD sum rules to the XAS and XMCD spectra, the effective spin ( m effspin = m spin + 7 m T ) andorbital magnetic moments have been deduced and plotted in Fig. 2. The hole number n h in the sumrules has been assumed to be 3.4 (Ref. 7). The figure shows that the orbital magnetic moment becomeshighly anisotropic with increasing S whereas the spin magnetic moments of Fe is nearly isotropic for S = 0 and 0 . . The spin magnetic moment of the S = 0 . sample is supposed to be isotropic, butthe in-plane XMCD spectrum of the S = 0 . sample could not be measured under a magneticallysaturated condition because the film with S = 0 . is not magnetically saturated up to about 10 T (seeSupplementary Materials Fig. S.1). For the same reason, the anisotropy of the the orbital magneticmoment of the S = 0 . sample is overestimated.Figures 3 shows the Pt L , -edge XAS and XMCD spectra of the L -FePt thin films ( S = 0, 0.4,3 = 90° H θ = 30° H FIG. 1. Fe L , -edge x-ray absorption spectroscopy (XAS) and x-ray magnetic circular dichroism (XMCD)spectra of L -ordered FePt thin films. (a)(b) XAS spectra for the positive (pink) and negative (green) photonhelicities for the out-of-plane (a) and in-plane (b) magnetic field directions. S denotes the degrees of long-rangechemical order. (c) Comparison of the XMCD spectra between the out-of-plane and in-plane magnetic fielddirections. All the spectra were taken at room temperature (RT). L and L edges are separated bya large energy of ∼ L and L edgeshave been normalized to 2.22 to 1 (Refs. 8 and 9). As shown in Figs. 3(c) and 3(d), the anisotropyof the XMCD intensity increases not only at the L edge but also at the L edge with S , in contrastto Fe L , -edge XMCD, where only the anisotropy of L -edge XMCD increases [see Fig. 1(c)]. Thespin and orbital magnetic moments of Pt estimated using the XMCD sum rules are plotted in Fig.4 as functions of S . The intensities of the XAS spectra, which appears in the XMCD sum rules ,have been estimated by the similar method as Refs. 8–10. First, the XAS spectra of a gold foil hasbeen subtracted from the raw XAS spectra of FePt so that the background above the main absorptionpeak vanishes. Then, assuming that the intensity of the subtracted spectra I diff is proportional to4he difference of the hole numbers of Pt ( n Pt h ) and Au ( n Au h ), we have deduced the XAS intensity ofPt I Pt as I Pt = n Pt h n Pt h − n Au h I diff The hole numbers n h have been assumed to be 1.73 for Pt and 0.75 forAu . The magnitudes of the spin and orbital magnetic moments are much smaller than those of Fe,as expected from the fact that Pt is a paramagnetic metal and its ferromagnetic moment is inducedthrough hybridization with the Fe 3 d orbitals. More significantly, the sign of the orbital momentanisotropy ∆ m orb (= m outorb − m inorb ) is opposite to that of Fe. Note that the in-plane XMCD spectrumof the S = 0 . sample could not be measured under a magnetically saturated condition with µ H = 7 T and, therefore, that the weak anisotropy of the spin magnetic moment of that sample would not beintrinsic. For this reason, the anisotropy of the orbital magnetic moment is slightly underestimated.Here we shall discuss the relationship between the MAE and the observed anisotropy of the orbitalmagnetic moment. Van der Laan has shown by treating the spin-orbit coupling as a perturbation thatthe MAE ( ∆ E ) are given by the following equations . ∆ E = − ξ µ B ˆ S · ( (cid:104) L ↓ (cid:105) hard − (cid:104) L ↓ (cid:105) easy )+ 212 ξ ∆ ex ˆ S · ( (cid:104) T (cid:105) hard − (cid:104) T (cid:105) easy ) , (1)where ξ is the spin-orbit coupling constant, (cid:104) L (cid:105) is the orbital magnetic moment, (cid:104) T (cid:105) is the magneticdipole term, and ∆ ex is the exchange energy. The first term in Eq. (1) originates from spin-conservedvirtual excitation in perturbation theory, and the second term from spin-flip virtual excitation. Brunohas derived a simplified relation between the MAE and the anisotropy of the orbital magnetic moment m spineff m orb FIG. 2. Effective spin magnetic moment m effspin (a) and the orbital magnetic moment m orb (b) of Fe in FePt filmderived from Fe L , -edge XMCD using the sum rules . The anisotropy of m effspin for the S = 0 . sample is dueto insufficient magnetic field of µ H = 5 T to saturate the in-plane magnetization. Likewise, the anisotropy of m orb for the same sample is overestimated to a similar extent. θ = 4.4° FIG. 3. Helicity-averaged XAS spectra of L -ordered FePt thin films at the Pt L (a) and Pt L edges (b) andthe XMCD spectra at the Pt L (c) and Pt L edges (d). The dashed curves are Au XAS spectra. by ignoring the second term of Eq. (1) : ∆ E = ξ µ B ( m easyorb − m hardorb ) . (2)Although the Bruno relationship [Eq. (2)] has been widely used, it may not be justified to ignore thecontribution of spin-flip term in the systems which contain elements with strong spin-orbit couplingsuch as Pt. Indeed, recent first-principles calculations on L -ordered FePt by Solovyev et al. andUeda et al. have shown that there is considerable contribution of Pt atoms as well as contributionof the spin-flip excitations to the MAE. The calculations have shown that the Pt sites contribute to theMAE through hybridization with the Fe 3 d orbitals .Table II summarizes the spin and orbital magnetic moments deduced from the present XMCD ex-periment and from the first-principles calculation by Solovyev et al. using the real-space Green’sfunction method. The calculated signs of ∆ m orb , positive for Fe and negative for Pt, agrees with theXMCD result, indicating that the first-principles calculation correctly predicts the magnetic proper-ties of L -ordered FePt. From the calculated MAE values decomposed into the Fe and Pt sites (thebottom entry of Table II), Solovyev et al. concluded that the MAE mainly originates from Pt. The6 spineff m orb FIG. 4. Effective spin magnetic moment m effspin (a) and the orbital magnetic moment m orb (b) of Pt derived fromPt L , -edge XMCD using the sum rules. The weak anisotropy of m effspin for the S = 0 . sample is due to theinsufficient magnetic field of µ H = 7 T to saturate the in-plane magnetization. Likewise, the anisotropy of m orb for the same sample is slightly underestimated. relationship between the anisotropy of the Fe 3 d orbital moment ( ∆ m orb > ) and the Fe 3 d compo-nent of the MAE ( < : in-plane easy magnetization axis) is opposite to that predicted by the Brunomodel. As for the Pt site, too, the anisotropy of the Pt 5 d orbital moment ( ∆ m orb < ) and the Pt 5 d component of the MAE ( > : perpendicular easy magnetization axis) is opposite to the Bruno model.Therefore, the observed anisotropy of the orbital moment indicates that the Bruno relationship doesnot hold even qualitatively for Pt. This means that the spin-flip term of Eq. (1), which representsthe magnetic dipole term arising from the anisotropic distribution of spin density and is ignored inthe Bruno model, should be important to discuss the origin of the MAE of L -ordered FePt. TheBruno model has been derived under the assumptions that the spin-orbit coupling is a weak pertur-bation compared to the exchange splitting and therefore that spin-flip term can be neglected. Theimportance of the spin-flip term for FePt arises from the large spin-orbit interaction and the smallspin splitting of the Pt 5 d electrons. Finally, as for the dependence of the spin and the anisotropy oforbital magnetic moments on the S , that is, the observation that m effspin decreases (increases) and ∆ m orb increases (decreases) for Fe (Pt) sites, respectively, as functions of S , is similar to the one deducedfrom the coherent-potential-approximation calculations by Staunron et al. and Kota et al. .In summary, we have performed XMCD studies of L -ordered FePt thin films with various S in order to reveal the relationship between the anisotropy of the orbital magnetic moments of Feand Pt and the magnetic anisotropy. Fe L , -edge XMCD studies have shown the existence of largeanisotropy of the orbital magnetic moment, that is, the orbital magnetic moment is larger for magneticfields perpendicular to the plane, which is enhanced with increasing S . Pt L , -edge XMCD studies7 ABLE II. Spin and orbital magnetic moments deduced from the present XMCD study [corrected for the insuf-ficient saturation for the in-plane magnetization so that m effspin does not depend on the magnetic field direction]and from the first-principles calculation by Solovyev et al. . The magnetic anisotropy energy (MAE) of theentire system deduced from magnetization measurements (see supplementary information) and those of indi-vidual elements deduced from the first-principles calculation are also indicated. Fe atom Pt atom m effspin m orb m effspin m orb Experiments( S = 0 . ) out-of-plane magnetic moment 2.28 0.21 0.30 0.036in-plane magnetic moment 0.12 0.043 ∆ m orb (= m outorb − m inorb ) – 0.11 – − (MJ / m ) m spin m orb m spin m orb First-principles calculation( S = 1 ) out-of-plane magnetic moment 2.77 0.0802 0.35 0.0486in-plane magnetic moment 0.0690 0.0616 ∆ m orb (= m outorb − m inorb ) – 0.0112 – − (MJ / m ) − show that the anisotropy of the orbital magnetic moment is opposite to that of Fe, that is, the orbitalmagnetic moment is larger for magnetic fields parallel to the plane than perpendicular to it. Thisresult is qualitatively consistent with the anisotropy of the calculated orbital magnetic moment fromfirst principles by Solovyev et al. , which calculation indicates the significant contribution of Pt andnegligible (or even opposite) contribution of Fe to the MAE of the entire FePt film. This is attributedto the strong spin-orbit coupling and small spin splitting of the Pt 5 d bands in L -ordered FePt. Thepresent result demonstrates that theoretical treatment including spin-flip processes is necessary whendesigning magnetic materials with large MAE utilizing heavy elements. SUPPLEMENTARY MATERIAL
See supplementary materials for the magnetization curves of each sample.
ACKNOWLEDGMENTS
This work was supported by Grants-in-Aid for Scientific Research from JSPS (grant Nos. 15H02109and 15K17696) and Nanotechnology Platform (project No. 12024046) from MEXT. The experimentwas performed at BL-16A of KEK-PF with the approval of the Photon Factory Program AdvisoryCommittee (proposal Nos. 2013S2-004, 2014G-177, 2016G066, and 2016S2-005) and at BL39XU ofSPring-8 with the approval of the Japan Synchrotron Radiation Research Institute (JASRI) (proposalNos. 2014A1158 and 2015B1461). A. F. is an adjunct member of Center for Spintronics Research8etwork (CSRN), the University of Tokyo, under Spintronics Research Network of Japan (Spin-RNJ). T. S. and K. T. are members of CSRN, Tohoku University, under Spin-RNJ. We thank V. K.Verma and T. Harano for technical support.
REFERENCES G. van der Laan, J. Phys.: Condens. Matter , 3239 (1998). P. Bruno, Phys. Rev. B , 865 (1989). T. Seki, H. Iwama, T. Shima, and K. Takanashi, J. Phys. D: Appl. Phys. , 335001 (2011). C. T. Chen, Y. U. Idzerda, H.-J. Lin, N. V. Smith, G. Meigs, E. Chaban, G. H. Ho, E. Pellegrin, andF. Sette, Phys. Rev. Lett. , 152 (1995). B. T. Thole, P. Carra, F. Sette, and G. van der Laan, Phys. Rev. Lett. , 1943 (1992). P. Carra, B. T. Thole, M. Altarelli, and X. Wang, Phys. Rev. Lett. , 694 (1993). C. Antoniak, M. E. Gruner, M. Spasova, A. V. Trunova, F. M. R¨omer, A. Warland, B. Krumme,K. Fauth, S. Sun, P. Entel, M. Farle, and H. Wende, Nat. Commun. , 528 (2011). W. Grange, M. Maret, J.-P. Kappler, J. Vogel, A. Fontaine, F. Petroff, G. Krill, A. Rogalev, J. Goulon,M. Finazzi, and N. B. Brookes, Phys. Rev. B , 6298 (1998). J. Bartolom´e, F. Bartolom´e, L. M. Garc´ıa, E. Roduner, Y. Akdogan, F. Wilhelm, and A. Rogalev,Phys. Rev. B , 014404 (2009). C. 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. , 117201 (2006). I. V. Solovyev, P. H. Dederichs, and I. Mertig, Phys. Rev. B , 13419 (1995). S. Ueda, M. Mizuguchi, Y. Miura, J. G. Kang, M. Shirai, and K. Takanashi, Appl. Phys. Lett. ,042404 (2016). J. B. Staunton, S. Ostanin, S. S. A. Razee, B. Gyorffy, L. Szunyogh, B. Ginatempo, and E. Bruno,J. Phys.: Condens. Matter , S5623 (2004). Y. Kota and A. Sakuma, J. Phys. Soc. Jpn.81