Magnetic polarization of copper in Cu-capped Co clusters

Magnetic polarization of copper in Cu-capped Co clusters

ARTICLE IN PRESS Journal of Magnetism and Magnetic Materials 316 (2007) e23–e26 www.elsevier.com/locate/jmmm Magnetic polarization of copper in Cu-c...

347KB Sizes 0 Downloads 32 Views

ARTICLE IN PRESS

Journal of Magnetism and Magnetic Materials 316 (2007) e23–e26 www.elsevier.com/locate/jmmm

Magnetic polarization of copper in Cu-capped Co clusters L.M. Garcı´ aa,, F. Bartolome´a, J. Bartolome´a, F. Luisa, F. Petroffb,c, C. Deranlotb,c, F. Wilhelmd, A. Rogalevd, P. Bencokd, N.B. Brookesd a

Instituto de Ciencia de Materiales de Arago´n, Dpto. Fı´sica de la Materia Condensada, CSIC-Universidad de Zaragoza, 50009 Zaragoza, Spain b Unite´ Mixte de Physique CNRS/Thales, Route De´partementale 128, 91767 Palaiseau Cedex, France c Universite´ Paris-Sud, 91405 Orsay Cedex, France d European Synchrotron Radiation Facility, BP220, F-38043 Grenoble, France Available online 25 February 2007

Abstract Magnetic anisotropy of Co nanoparticles homogeneously dispersed in alumina is increased by capping the particles with copper. In the present work, we investigate this effect by measuring X-ray magnetic circular dichroism (XMCD) at the K and L2,3 edges of Co and Cu atoms. We demonstrate that orbital-to-spin ratio at the Co 3d states is increased by capping with Cu, explaining the observed enhancement of the macroscopic anisotropy. Hybridization between Co and Cu at the interface produces an electronic transfer between them. Clear induced magnetic moments are observed in copper, both in the 4p states and in the 3d states. r 2007 Elsevier B.V. All rights reserved. PACS: 61.46.Df; 61.10.Ht; 75.75.+a Keywords: Magnetic nanoparticle; XMCD; Magnetic anisotropy; Copper magnetism

Magnetic nanoparticles are mesoscopic materials of enormous interest as they provide the opportunity to investigate the evolution of magnetic properties from isolated atoms towards the bulk and due to their possible technological applications. It is well known that the magnetic anisotropy K of metal particles is strongly enhanced with respect to the bulk values [1–3], as also happens in magnetic films. The enhancement of K (one of the key parameters for applications) has been attributed to the incomplete quenching of the orbital magnetic moments mL at the surface [3–5]. In a previous work [6], we have systematically studied the evolution of K of nanometric Co clusters (from 1 to 5 nm) embedded in alumina with the cluster size, showing that K is dominated by the strong anisotropy induced at the surface of the clusters. In metallic films, the magnetic anisotropy depends not only on the layer’s thickness but also on the nature of the substrate on which the films grow [7], due to the bonding of surface magnetic atoms to those of the substrate or capping Corresponding author. Tel.: +34 976 762456; fax: +34 976 761229.

E-mail address: [email protected] (L.M. Garcı´ a). 0304-8853/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jmmm.2007.02.016

layers. In this line, we have reported that capping Co clusters by a thin metal noble layer like Cu [8] or Au [9], also provides a simple method to enlarge i.e. to tailor, the anisotropy of a granular system: surface anisotropy increases up to a factor 3 by capping with Au with respect to the particles in an alumina matrix. The objective of the present work is to investigate by means of an element and shell selective technique like X-ray magnetic circular dichroism (XMCD) the effect of capping the Co clusters with a Cu layer. Hybridization at the interface between Co and Cu should change the electronic structure and magnetism of both. Indeed spin polarization of Cu in Co/Cu multilayers has been detected both at the L2,3 edges of Cu (3d states) [10] and at the K edge (4sp hybridized bands) [11,12], and very recently in the 3d states in CoCu granular alloys [13]. Nanometer-sized Co clusters were prepared, as described previously [14,15], by sequential sputtering of amorphous alumina and Co layers on an Si substrate. The Co layer segregates into nearly spherical metallic clusters, which within each layer are separated by an average distance of 2.2 nm between particle surfaces [15]. Their size can be

ARTICLE IN PRESS L.M. Garcı´a et al. / Journal of Magnetism and Magnetic Materials 316 (2007) e23–e26

e24

controlled by varying the amount of Co deposited by layer. This amount is given by the nominal thickness tCo that the Co film would have if it were continuous. Below tCo ¼ 1 nm, the clusters crystallize with the fcc structure. In this work, we studied granular films with tAl2 O3 ¼ 3 nm and tCo ¼ 0.7 nm, corresponding to an average Co particle diameter of 3 nm as determined from TEM and magnetic measurements [6,14,15]. Copper-capped clusters were prepared by depositing a continuous thin (1.5 nm) Cu film onto the previously formed Co clusters. Therefore, we do not expect the crystal structure and morphology of the clusters to be affected by the capping, as indeed has been tested by magnetic measurements. The two samples (capped and uncapped) are made by piling up 15 repetition units of Al2O3/Co or Al2O3/Co/Cu. We performed two kinds of XMCD experiments at the ESRF in Grenoble. Hard X-ray experiments (K edge of Co and Cu) were performed at the ID12 beamline using total fluorescence yield detection in backscattering geometry. In addition, soft X-ray XMCD measurements (L2,3 edges of Co and Cu) were performed at the ID08 beamline using total electron yield detection. All experiments were made at 5 K, under magnetic field of 1 T (enough to magnetically saturate the sample), and reversing both helicity of the incident X-rays and direction of the applied magnetic field. All the X-ray absorption spectra (XAS) were normalized at high energies, after background subtraction, to remove thickness dependence. In all cases, the origin of the energy scale was chosen at the inflection point of the absorption edge. Subtraction of the normalized XAS spectra (parallel and antiparallel alignment of helicity and magnetization) gives us the XMCD signal. Fig. 1 shows the normalized XMCD signal at the Co K edge of the Al2O3/Co and Al2O3/Co/Cu samples. These signals resemble both in shape and in magnitude the typical K edge XMCD signal of metallic Co [11,12]. The magnetic

moment on the 4p band of Co (0.072(4) mB per atom) is created by the spin-dependent polarization of the exchange polarized 3d band. The hybridization tends to align the p moment antiparallel to the 3d moment. The XMCD amplitude of the Cu-capped sample is slightly smaller (10%) than the amplitude of the uncapped one. This effect could be related with the observed reduction of total Co moment in Cu-capped Co films [16] with respect to the uncapped ones, although in our case the relative decrease of the moment is significantly smaller. In Fig. 2, we show the normalized XMCD signal at the Co L2,3 edges for the same samples. It is clearly visible that the addition of Cu increases the ratio between the L3 and the L2 XMCD peaks of Co. Applying the XMCD sum rules [17], we find that the ratio mL/mS of orbital-to-spin magnetic moments increases from 0.13(1) for the uncapped clusters to 0.16(1) for the Cu-capped clusters, both larger than the bulk 0.08(1) [18]. This suggests that the Co–Cu interaction at the interface changes the occupation of Co 3d orbitals. This is evidenced further in the inset of Fig. 2, showing the integrated XAS signals for both samples. Since these integrals are directly related to the number of holes of the Co 3d states [17,18], the observed reduction in the Cucapped sample with respect to the uncapped one indicates that capping with copper increases the number of 3d electrons in Co. The enhancement of the Co orbital moment by capping with Cu had been previously observed in Co/Cu interfaces [5] and agrees well with theoretical calculations [19]. This orbital-to-spin ratio enlargement is at the origin of the observed enhancement of K by capping with Cu [8]. Sum rules yield a total 3d magnetic moment of Co of 1.66 mB per atom, in good agreement with the value found in thin films [18]. No significant differences, within the limits of the experimental resolution, are observed between the total 3d moments of the uncapped and the Cu-capped samples.

Co L2,3 edges

0.002

5K, 1T

Co K - edges 5 K, 1 T

0

0.000

uncapped Integrated XAS (a.u.)

XMCD (a.u.)

XMCD (a.u.)

0.001

-1

-0.001 -2 -0.002

Cu - capped

uncapped Cu - capped

E - E0 (eV)

Cu - capped

uncapped

-0.003 -20

-20 0

20 E-Eo(eV)

40

60

Fig. 1. Normalized XMCD spectra at the Co K edge obtained at T ¼ 5 K and applied magnetic field of 1 T of Al2O3/Co and Al2O3/Co/Cu multilayers.

-10

0

10 E-E0 (eV)

20

30

40

Fig. 2. Normalized XMCD spectra at the Co L2,3 edges obtained at T ¼ 5 K and applied magnetic field of 1 T of Al2O3/Co and Al2O3/Co/Cu multilayers. The inset shows the energy dependence of the integrated XAS signal for the same samples.

ARTICLE IN PRESS L.M. Garcı´a et al. / Journal of Magnetism and Magnetic Materials 316 (2007) e23–e26

1.2 1.0

Cu K - edge 5K, 1 T

Absorption (a.u.)

0.8 0.6 0.4 XMCD x 1000

0.2

1.0

Cu L2,3 edges 5K, 1T

0.8 Absorption (a.u)

Evidence for electronic interaction at the Co/Cu interface is also observed in the absorption spectra of Cu. In Fig. 3, XAS and XMCD spectra at the K edge of Cu for the Al2O3/Co/Cu sample are presented. The XMCD spectrum has the same sign than that of Co (Fig. 1), indicating a ferromagnetic polarization of Cu sp states (responsible of conduction) by Co moments. The signal is also similar to that observed in Co/Cu multilayers [11,12] and scaling with XMCD data of Ref. [11] gives an average magnetic moment of 0.016(1) mB per atom (i.e. antiparallel to 3d Co moment). This is the first time that spin polarization at the sp levels of Cu is experimentally observed in granular systems, although its existence had been theoretically predicted [20]. Fig. 4 shows the absorption and XMCD spectra at the L2,3 edges of Cu for the Al2O3/Co/Cu sample. First, it is interesting to note that the L3 and L2 white lines are clearly enhanced with respect to a Cu foil, for instance shown in Ref. [21]. This points towards an increase in the number of 3d holes at the Cu atoms. These electrons have been transferred to the Co, as discussed above. Second, a clear XMCD signal is observed with the same sign as that of the Co XMCD peak, as shown in Fig. 2. This shows that 3d Cu electrons are ferromagnetically polarized by Co. Comparison with the signal observed in Co/Cu multilayers gives a 3d magnetic moment for Cu of around 0.023(3) mB per atom. It is interesting to note that the Cu magnetic moments on the d band and on the p band have opposite signs and similar magnitudes (around 0.02 mB). This does not necessarily imply that the total magnetic polarization of Cu vanishes, because the polarization of the Cu d states decays faster with Cu thickness than the polarization of the p states. Indeed, for 3d states, it is almost limited to the Co/Cu interface [10], whereas for p states it extends to the outer Cu atomic layers [11,22]. Theoretical calculations [10] give a Cu polarization at the Co/Cu surface 30% higher for the d states than for the sum of the s and p states. As

e25

0.6 0.4 0.2 XMCD x 20

0.0 -0.2 -0.4 -10

0

10 20 E- Eo (eV)

30

40

50

Fig. 4. Normalized absorption and XMCD spectra at the Cu L2,3 edges obtained at T ¼ 5 K and applied magnetic field of 1 T of Al2O3/Co/Cu multilayers. The latter has been multiplied by 20.

XMCD measurements average the signal of the whole Cu layer, the lower magnitude of sp states is compensated by the slower decay along the Cu thickness, giving rise to similar experimental values. Application of sum rules [17] to the XMCD performed at the L2,3 edges of Cu enables us to determine a ratio mL/ mS of around 0.20(2). This ratio is similar to that observed in the Co atoms. This result differs clearly from that obtained from XMCD measurements on Co/Cu multilayers [10] which does not detect any orbital moment, assigning a spin character to the Cu 3d moment. However, large orbital moments have also been detected at the Cu 3d states in FeCu superlattices [21]. Summarizing, XMCD experiments clearly indicate that capping the Co nanoclusters with Cu increases the orbitalto-spin ratio at the Co 3d states. This result explains the observed enhancement of the macroscopic anisotropy. No evident effects are observed in the 3d Co moment, whereas a small reduction is observed in the moment of the 4p band. Hybridization between Co and Cu at the interface produces an electronic transfer between them. Clear induced magnetic moments are observed in copper, both in the 4sp states and in the 3d states. They are parallel to the respective ones of Co (ferromagnetic polarization), antiparallel between them, and they have similar magnitudes. Our results in granular systems agree well with the reported ones in multilayers and are supported by theoretical calculations.

0.0

Financial support from Spanish CICyT (projects MAT0501272 and MAT05-02454) and DGA is acknowledged.

-0.2 -0.4 -20

-10

0

10

20

30

40

50

E-Eo (eV) Fig. 3. Normalized absorption and XMCD spectra at the Cu K edge obtained at T ¼ 5 K and applied magnetic field of 1 T of Al2O3/Co/Cu multilayers. The latter has been multiplied by 1000.

References [1] J.P. Chen, et al., Phys. Rev. B 51 (1995) 11527. [2] M. Respaud, et al., Phys Rev. B 57 (1998) 2925. [3] P. Gambardella, et al., Science 300 (2003) 1130.

ARTICLE IN PRESS e26 [4] [5] [6] [7] [8] [9] [10] [11] [12] [13]

L.M. Garcı´a et al. / Journal of Magnetism and Magnetic Materials 316 (2007) e23–e26 P. Bruno, Phys Rev. B 39 (1989) R865. M. Tischer, et al., Phys. Rev. Lett. 75 (1995) 1602. F. Luis, et al., Phys Rev. B 65 (2002) 094409. M. Kisielewski, et al., Phys. Rev. Lett. 89 (2002) 087203. F. Bartolome´, et al., J. Magn. Magn. Mater. 272–276 (2004) e1275. F. Luis, et al., J. Appl. Phys. 99 (2006) 08G705. M.G. Samant, et al., Phys. Rev. Lett. 72 (1994) 1112. S. Pizzini, et al., Phys. Rev. Lett. 74 (1995) 1470. S. Nagamatsu, et al., Phys Rev. B 70 (2004) 174442. A. Garcı´ a Prieto, et al., Phys Rev. B 72 (2005) 212403.

[14] [15] [16] [17] [18] [19] [20] [21] [22]

J. L Maurice, et al., Philos. Mag. A 79 (1999) 2921. J. Bria´tico, et al., Eur. Phys. J. D 9 (1999) 517. A. Ney, et al., Europhys. Lett. 54 (2001) 820. B.T. Thole, et al., Phys. Rev. Lett. 68 (1992) 1943; P. Carra, et al., Phys. Rev. Lett. 70 (1993) 694. C.T. Chen, et al., Phys. Rev. Lett. 75 (1995) 152. O. Hjorststam, et al., Phys Rev. B 53 (1996) 9204. X. Chuanyun, et al., Phys Rev. B 55 (1997) 3677. W. Kuch, et al., Phys Rev. B 58 (1998) 8556. K. Hirai, Physica B 345 (2004) 209.