Spin and orbital moment in amorphous Co68Fe24Zr8 layersThomas Hase,1 Hossein Raanaei,2, 3 Hans Lidbaum,4 Cecilia Sánchez-Hanke,5Stuart Wilkins,6 Klaus Leifer,4 and Björgvin Hjörvarsson2, ∗1Department of Physics, University of Warwick, Coventry, CV4 7AL, United Kingdom2Department of Physics and Materials Science, Uppsala University, Box 530, 751 21 Uppsala, Sweden3Department of Physics, Persian Gulf University, Bushehr 75168, Iran4Institute of Electron Microscopy and Nano-Engineering,Uppsala University, Box 534, SE 751 21, Sweden5National Synchrotron Light Source, Brookhaven National Laboratory, Upton, New York 11973, USA6Department of CMPMSD, Brookhaven National Laboratory, Upton, New York 11973, USA(Dated: July 13, 2009)The ratio of the orbital to the spin magnetic moment was determined for both Fe and Co in amor-phous Co68Fe24Zr8 layers using x-ray circular dichroism. The investigations were performed on boththick Co68Fe24Zr8 layers as well as on amorphous Co68Fe24Zr8/Al70Zr30 multilayers grown by DCsputtering. Structural characterization was performed using x-ray re�ectometry, x-ray di�ractionand transmission electron microscopy. X-ray circular dichroism, x-ray magnetic scattering as wellas the magneto-optic Kerr e�ect were used to characterise the magnetic properties of the amor-phous materials. The ratio of the orbital to spin moments in the single CoFeZr layer sample was0.012 ± 0.005 for Fe and 0.078 ± 0.005 for Co. Substantial reduction in the the ratio of the orbitalto spin moments was observed with decreasing CoFeZr layer thickness.PACS numbers:I. INTRODUCTIONThe physical properties of amorphous materials arecurrently only partially understood and there are cur-rently no theoretical calculations which can be used topredict their properties from �rst principles. Althoughamorphous materials are relatively common, most oftheir physical properties are much less known thantheir crystalline counterparts. The magnetic propertiesare no exception, although amorphous magnetic mate-rials are widely used1,2. Furthermore, there exists evencounter intuitive properties which are often utilized butpoorly understood. One such example is the �eld in-duced anisotropy of amorphous magnets3,4. The localanisotropy, seen as a spatially varying anisotropy in thinamorphous �lms, can be argued to have the same ori-gin. The subsequent control of this induced magneticanisotropy is critical in most applications including spin-electronics.The relationship between the spin and orbital compo-nents of the magnetic moment are fundamental for theunderstanding of both the macroscopic and microscopicproperties of magnetic materials. Spin-orbit coupling canbe viewed as the source of magnetic anisotropy and so thelocal symmetry of the atomic arrangements plays an im-portant role in determining the observed anisotropy. Theatomic arrangement within a material is often linked in-timately to the underlying lattice in a crystal structureand is re�ected in the term crystalline anisotropy, whichis one of the basic parameters used to describe magneticmaterials. In amorphous materials magneto-crystallineanisotropy must be absent due to the lack of long-rangeorder but a macroscopic magnetic anisotropy is still ob-served. Although there is no long-range order, variation

in the local arrangement of atoms on the nanometer rangeappears to be su�cient to break the local symmetry andthereby give rise to the anisotropy observed in amorphousmaterials5,6. Thus, although the long range crystallineorder is absent, the local atomic arrangement allows forthe creation of easy and hard magnetic axes within amor-phous magnetic materials which can be further modi�edthrough for example ion implantation7.The orbital moments in bulk materials of high symme-try within crystalline solids are strongly reduced due toorbital quenching. However, they may attain large val-ues when the dimensionality is reduced. In contrast tocrystalline materials, thin �lms of amorphous magneticalloys show large variations in both the spin and angularmomentum as a function of both composition and dimen-sionality. Thus, a detailed understanding of the inter-play between the magnetic and structural properties ofsolids is required for the tailoring of magnetic propertiessuch as interface magnetism in new arti�cial amorphousmagnetic systems8.X-ray magnetic circular dichroism (XMCD) allows theelement speci�c determination of the orbital and spinmagnetic moments 9,10 and thereby provides a directroute for exploring the element speci�c contribution tothe magnetic properties in materials11�15. In this pa-per we address the orbit to spin ratio (morbit / mspin)of the magnetic moment of both Fe and Co in amor-phous CoFeZr magnetic layers using XMCD. The thick-ness dependence of morbit / mspin is also addressed andthe interface contribution as well as the element speci�ccontributions to the magnetic properties of CoFeZr arediscussed.

2 II. EXPERIMENTA series of samples were grown using DC mag-netron sputtering on naturally oxidized Si(111) sub-strates, which were annealed at 530◦C for one hour priorto deposition. The base pressure of the chamber was be-low 2 x 10−9 Torr. During the deposition of the layers,Ar gas with a purity of 99.9999% was kept at a pressureof 3.0 x 10−3 Torr. CoFeZr layers were deposited froma single composite target with a nominal composition ofCo60Fe25Zr15 whilst the Al70Zr30 layers were grown byco-deposition from two elemental sources. All the sam-ples were grown on a nominally 30 Å thick Al70Zr30 bu�erlayer at room temperature and were capped by aluminumto prevent oxidation of the magnetically active layers.Here we discuss the results from two sets of sam-ples. The �rst sample contains a single CoFeZr layerand has a nominal structure: Si/SiO2/CoFeZr(200 Å).We compare this single, thick layer with a sec-ond series of multilayer samples consisting of 10repeats of a bilayer with varying CoFeZr thick-ness. The nominal structure of the multilayers are:Si/SiO2/AlZr(30 Å)/[CoFeZr(X Å)/AlZr(30 Å)]×10,with X=10, 20 and 30 Å these are labelled sample 10/30,20/30 and 30/30 respectively. For both sets of samplesa twenty Ångstrom aluminum capping layer protects theunderlying magnetic material by forming a passivatinglayer of oxidized aluminum. The chemical compositionof the samples was con�rmed using Rutherford Backscat-tering Spectrometry whilst the amorphous nature of thesamples was investigated by using X-ray di�raction andTransmission Electron Microscopy (TEM). For the highresolution TEM studies a FEI Tecnai F30ST microscopeoperating at 300 kV was used. Electron di�raction wasperformed on a JEOL JEM-2000FXII instrument usingan electron beam energy of 200 kV.Layer and interface parameters were determined us-ing grazing incidence scattering techniques. Specular re-�ectivity and di�use scattering were recorded on stationX22C at the National Synchrotron Light Source (NSLS)at Brookhaven National Laboratory. A silicon (111)monochromator was used to provide a beam of energy8.8 keV which was focused vertically onto the sample us-ing a mirror and slits with dimensions 150 µm × 1 mm.The scattered radiation was measured using a slit colli-mated scintillator detector with a resolution matched tothe incident beam.The bulk magnetization of the samples was determinedthough Magneto-Optical Kerr E�ect (MOKE) measure-ments recorded at room temperature using a longitudi-nal geometry with polarized light. Element speci�c mag-netic information was recorded using XMCD and reso-nant magnetic scattering (XRMS) on beam-line X13Aat the NSLS. An elliptical polarizing wiggler producesradiation covering the L edges of the transition metalsof interest. The polarization of the incident light wasswitched at 22 Hz and a phase-sensitive detection sys-tem using lock-ins was employed to record simultane-

ously the magnetic and charge signals by measuring at22 and 44 Hz respectively. The sample was mounted ina high vacuum two-circle di�ractometer and the experi-ment was conducted in a horizontal scattering geometry.The sample was magnetized in-situ using an electromag-net with a maximum �eld of 20 mT applied parallel to thesample surface. The drain current from the sample wasmeasured by a�xing a thin wire to the sample surface.The re�ected intensity was recorded using an avalanchephotodiode (APD) with a digital counting chain synchro-nized to the helicity reversals.III. RESULTSA. Structural and compositional characterizationIn this section we elucidate the layer thicknesses andinterface morphologies using grazing incidence re�ectiv-ity. The q-vectors that are probed in such a geometryare insensitive to the crystal structure of the material.Representative re�ectivity data is shown in Fig. 1 for thesingle CoFeZr �lm. The forward di�use scatter has beenremoved prior to �tting by subtracting the longitudinaldi�use data which was recorded with an o�set in thesample angle of −0.1◦ from the specular condition. There�ectivity has been �tted using a genetic algorithm tominimize the di�erence between the data and a modelstructure using the GenX code16. Excellent agreementbetween simulation and data has been achieved with theperiodicity and amplitude of the re�ectivity fringes be-ing well reproduced in the simulation. The �tted CoFeZrlayer thickness is close to the nominal value, 202.0±0.1 Åwith an interface width of 4.7±0.1Å. The aluminum cap-ping layer was deduced to be close to fully oxidized witha �tted thickness of 27.3± 0.2 Å and a surface roughness5.1 ± 0.2 Å identical to both the substrate (4.6 ± 0.4 Å)and CoFeZr layers. The relative concentration of theconstituents of the CoFeZr layer was determined fromRBS measurements which yielded a layer compositionof Co68Fe24Zr8 which was in excellent agreement withthe re�ned CoFeZr electron density determined from thespecular re�ectivity �t.The longitudinal di�use data shown in Fig. 1 showsfringes with the same periodicity as those found in thespecular data. This is indicative of a high degree of ver-tical correlation in the roughness which extends from thesubstrate through to the sample surface. As the ampli-tude of the interface width for all interfaces is similar,we can conclude that much of the topological roughnessis correlated. The amplitude of the topological rough-ness can be estimated by considering the ratio of theintegrated specular to di�use scatter in a transverse dif-fuse scan. Such a scan recorded at qz = 0.222 Å−1is shown in Fig. 2 and a lower-limit on the amplitudeof the topological roughness is 1.0 ± 0.2 Å. From therather large discrepancy between the specular and di�useanalysis we conclude that the interface contains signi�-

310-3

10-1

101

103

105

0 0.2 0.4 0.6

qz(Å-1)

Nor

mal

ised

Inte

nsity

(arb

.uni

ts)

Figure 1: (Color Online) Specular re�ectivity (upper curve)and longitudinal di�use scan (lower curve) for the single 200 ÅCoFeZr layer. The specular data (circles) are well �tted (line)for all scattering vectors. The fringes observed in the speculardata are replicated in the longitudinal di�use data recordedwith a sample o�set of −0.1◦, clearly illustrating that verticalcorrelations exist in the topological roughness that extendover the entire sample thickness.0.001

0.01

0.1

1

10

100

-6 -4 -2 0 2 4 6

qx(x 10-3 Å-1)

Nor

mal

ised

Inte

nsity

(arb

.uni

ts)

Figure 2: (Color Online) Transverse di�use, or rockingcurve, for the 200 Å single layer CoFeZr sample. The datawere recorded at a �xed out-of-plane scattering vector ofqz = 0.222 Å−1 . The specular ridge at qx = 0 is broadenedas a result of roughness on a large lateral length scale. Fromthe ratio of the specular and di�use scatter a lower-limit forthe amplitude of the topological roughness is 1.0 ± 0.2Å.cant inter-di�usion between the layers. This is in agree-ment with our previous study of similar amorphous ma-terial4. It is worth noting that the width of the spec-ular ridge at qx = 0 in Fig. 2 is considerably broader((2.5 ± 0.3) × 10−4 Å−1

) than the instrumental resolu-tion ((3.2 ± 0.1) × 10−5 Å−1

). This shows that thereexist some large in-plane structural features that are be-yond the coherence length of the x-ray beam (which atthis scattering vector is approximately 20µm). These aremost likely to be associated with large terraces commonlyobserved on Si(111) substrates.Incorporating the CoFeZr amorphous materials into

10-3

100

103

106

109

0 0.2 0.4 0.6

(a)

(b)

(c)

qz(Å-1)

Nor

mal

ised

Inte

nsity

(arb

.uni

ts)

Figure 3: Specular re�ectivity for the multilayer samples. Se-quential curves have been o�-set by decades of intensity forclarity. The multilayer peak positions correspond to �tted bi-layer thicknesses of 39.3±0.1 Å, 47.9±0.1 Å, 57.3±0.5 Å forsamples 10/30, 20/30 and 30/30 respectively. The solid linesare simulated �ts to the data (circles).magnetic heterostructures composed of repeated bilay-ers of thin CoFeZr and amorphous AlZr does not signif-icantly alter the structural parameters. The re�ectivitycurves for samples 10/30, 20/30 and 30/30 with increas-ing CoFeZr thickness are shown in Fig. 3. A summary ofthe layer thicknesses and roughness derived from the �tsare presented in Table. I.As in the case of the single CoFeZr �lm, the layerthicknesses are close to the nominal values and the in-terface widths are low. We note that the roughness ofthe AlZr/CoFeZr and CoFeZr/AlZr interfaces are asym-metric. The roughness of the CoFeZr/AlZr interface de-creases somewhat with increasing CoFeZr thickness, seeTable. I. As was the case for the single layer samplefringes are observed in the longitudinal di�use scans thatmatch features in the specular scans are observed for allmultilayer samples clearly demonstrating that the topo-logical roughness is strongly correlated between the lay-ers. The length scales over which these correlations ex-tend can be extracted from a full reciprocal space mapof the di�use data which is shown for the sample 20/30in Fig. 4. The di�use data is clearly con�ned to strongstreaks of di�use scatter that occur at the out-of-planereciprocal space, qz, positions corresponding to the low-angle multilayer peaks seen in Fig. 3. The broad ex-tension in qx of the di�use Bragg sheets shows that thefrequency components of the roughness that are corre-lated over a single bilayer extend down to a length scalesless than 500 Å. The slight curvature of the Bragg sheetsat high qx is caused by refraction e�ects when either theincident or exit beams correspond to the critical angle.Apart from the broadened specular ridge that was seenin Fig. 2, the transverse rocking curves recorded at theBragg peaks are rather �at (consistent with a short in-plane correlation length) and give a lower limit for thecorrelated topological roughness of 0.8± 0.2Å in excellent

4 Parameter Sample 10/30 (Å) Sample 20/30 (Å) Sample 30/30 (Å)Bilayer thickness 39.3 ± 0.1 47.9 ± 0.1 57.3 ± 0.5AlZr 29.8 ± 0.5 28.3 ± 0.5 29.8 ± 0.2CoFeZr roughness 5.5 ± 0.1 6.09 ± 0.08 7.8 ± 0.2AlZr roughness 11 ± 1 7.6 ± 0.3 6.3 ± 0.1Table I: Average bilayer thicknesses and AlZr thicknesses of the sample 10/30, 20/30 and 30/30 accompanied by AlZr andCoFeZr roughnesses are determined from simulations of the X-ray re�ectivity curves.agreement with the single layer sample. The strong cor-relation of the interface morphology between the layerscan also be seen in the HRTEM images shown in Fig. 5where the waviness of the layers is well replicated fromthe substrate to the surface. A clear example is seenby a typical cusp can be seen in Fig. 5b, indicated bya white arrow. Note how the cusp causes a change inthe interface morphology which subsequently propagatesthroughout the remainder of the stack while preservingthe bilayer thickness between the layers.

-0.01 -0.005 0 0.005 0.01

0.05

0.1

0.15

0.2

0.25

0.3

-5

-2

1

4

7

In-plane reciprocal lattice vector, ( )qxC

-1

Out-

of-

pla

ne r

eci

pro

cal l

attic

e v

ect

or,

()

qzC

-1

Figure 4: (Color Online) Full reciprocal space map of the dif-fuse scattering data from sample 20/30. Data is presented ona logarithmic scale. The di�use scattering is predominantlyfound in streaks at the out-of-plane scattering vectors of themultilayer peaks seen in the specular scans shown in Fig. 3.The slight curvature of the Bragg sheets at high qx is causedby refraction e�ects when either the incident or exit beamsare at the critical angle.Di�raction and microscopy studies con�rm the amor-phous structure of both the CoFeZr and AlZr layers.High resolution TEM images of the thinnest, 10/30, andthickest multilayer sample, 30/30, in cross section areshown in Fig. 5a and Fig. 5b, respectively. The siliconsubstrates were oriented close to the [110] zone axis in or-der to have the electron beam parallel with the depositedlayers. Both samples exhibit amorphous multilayer struc-ture, although with some interface inter-mixing betweenthe layers as well as correlated waviness. Note that the

aluminum cap layer of the thinnest sample shows crys-talline lattice fringes, thus the capping is not fully oxi-dized in this sample.

Figure 5: High resolution TEM images of the CoFeZr/AlZrmultilayer samples in cross sectional view. The sample withthe thinnest CoFeZr layer thickness, 10/30 is shown in a). Anamorphous structure is observed throughout the depositedmaterial, except for the Al cap layer at the top. In b) thesample with the thickest CoFeZr layer, 30/30, is shown. Thewhite arrow indicates a region where the interface morphologyis changed which then propagates throughout the remainderof the multilayers. The silicon substrate is seen at the bottomof each image.To further con�rm the amorphous structure of the lay-ers, selected area electron di�raction patterns (SAED)

5

Figure 6: Selected area electron di�raction pattern in theTEM from the 10 Å CoFeZr multilayer sample in top viewgeometry, acquired including a part of the silicon substrate(spot pattern). Broad dark bands indicates an amorphousstructure of the CoFeZr and AlZr layers. The sharp ring atapproximately 1.44 Å correspond to polycrystalline Al 220planes, originating from the cap layer (see Fig. 5). White ar-rows indicate the expected spacing of (111), (200), (220) and(113) re�ections in Al. The di�raction pattern was invertedfor clarity.were acquired in top view sample geometry. The siliconsubstrate was oriented to have the electron beam in the[111] zone axis direction and illuminating a region of ap-proximately 2.1µm in diameter. A SAED pattern of thethinnest sample, 10/30 is shown in Fig. 6. The well de-�ned spot pattern originates from the silicon substratewhile the broad rings come from the amorphous materialwithin the sample. At a di�raction angle correspondingto a lattice spacing of approximately 1.44 Å a sharp ringis observed which corresponds to the (220) atomic planesin face centered cubic aluminum, which from Fig. 5 werealso was found in the cap layer. The intensity of thecrystalline cap layer is barely visible since the di�ract-ing volume is small and will be superimposed with theamorphous bands of the CoFeZr and AlZr. The expectedpositions of the (111), (200), (220) and (113) re�ectionsin Al are indicated with white arrows in Fig. 6. Thecomposition of the magnetic material was determined byRutherford Back Scattering and is reported in more de-tail in our previous work4 and is Co68Fe24Zr8.B. Magnetic characterizationThe �eld dependence of the magnetization for the200 Å thick sample is shown in Fig. 7. The measure-ments were performed at room temperature, using theMagneto Optical Kerr E�ect. As seen in �gure 7, the

sample appear to be ferromagnetic, with a coercivity ofabout 1 mT. After some rounding at low �elds the sam-ple is fully saturated by about 5 mT. The sample ismagnetically isotropic in the plane as both the satura-tion and coercive �elds are independent of the alignmentof the sample in the applied �eld. The reduction in themagnetization at the lowest �elds originates from the ran-domness in the local anisotropy of the layers. The e�ectof the randomness in the local anisotropy can be seenas wandering axes of the direction of the magnetizationwhich has been observed by magnetic imaging for �eldsbelow the saturation value 4.

-10 -8 -6 -4 -2 0 2 4 6 8 10-1

0

1

Kerr

rota

tion

(arb

.uni

ts)

Field (mT)Figure 7: Room temperature magnetization as a function ofapplied �eld recorded using the Magneto-optical Kerr rotatione�ect.-1

0

1

-10 -5 0 5 10

Field (mT)

Mag

netis

atio

n(M

/MS)

Figure 8: (Color Online) The di�erence in the re�ected signalas a function of applied �eld for the single layer sample. TheFe (cirles and line) and Co (line) loops are similar withinexperimental error and show a coercive �eld of 1.7± 0.1 mT.The element speci�c magnetization was obtained usingresonant X-rays. Unlike optical probes which interrogatethe average magnetic moment within their sample depth,resonant X-rays can be used to extract the element spe-ci�c magnetic response when tuned to an appropriateenergy. X-ray absorption (XAS) and magnetic dichroism

6(XMCD) have become standard tools in characterizingmagnetic materials. For the transition materials the res-onant transition of interest occurs at the L3 and L2 edgesbecause the absorption is between the 2p core level andthe unoccupied, but spin-polarized 3d states.Both Fe and Co loops recorded in re�ection are identi-cal and are illustrated in Fig. 8. The Fe and Co are there-fore ferromagnetically aligned and rotate coherently withthe applied �eld. The noise on the Fe signal is larger thanthat observed from the Co due to the signi�cantly lowerFe concentration. The coercive �eld obtained from thex-ray measurements is somewhat larger than obtainedfrom the MOKE measurements which can be traced tothe remnant �eld in the magnet used in the X-ray mea-surements. The obtained saturation �eld is in excellentagreement with that obtained by MOKE, about 5 mT.At this �eld the importance of the slight o�set from theremnant �eld from the magnet has diminished. Thusthe MOKE and the XRMS measurements are consistent,when considering the experimental uncertainties.The absorption of circular polarized light that givesrise to the XMCD signal obeys the so-called sum rules,derived assuming electric dipole transitions only. The or-bital component of the magnetization contributes withthe same sign to the XMCD signal in a ratio 2:1 atthe L3 and L2 edges respectively. The spin componentarises from the core-hole splitting of the 2p states dueto the spin-orbit interaction and the 3d spin momentcontributes equally to the two edges but with oppositesign. Thus by considering the integrals under the XMCDcurves at the two edges, the spin and orbital componentsof the magnetic moment, can in principal be determined:mspin = −

1

C(A − 2B)µB, (1a)

morbit = −2

3C(A + B)µB . (1b)In Eq. 1a and 1b A is the integral of the XMCDsignal across the L3 edge and similarly B is the integralacross the L2 edge. The constant C is element speci�cand contains the number of holes per unit area and theexact polarization of the incident photons.The absorption spectra were recorded using the elec-tron yield method in a �xed, saturating �eld of ±10 mTfrom a wire attached to the surface of the �lm. Data wasrecorded using a lock-in measuring at twice the helicityswitching frequency at a sample angle of 30◦. In order toremove artifacts introduced by the applied �eld the datafrom the two �elds was averaged. The resulting normal-ized XAS spectra across the L3 and L2 edges of Fe andCo for the thick single �lm are shown in Fig. 9. Follow-ing standard procedures the background in the pre-edgeregion has been removed and the step height normalizedto unity well above the L2 edge. In Fig. 10 we show thedi�erence data which was recorded simultaneously withthe absorption data by measuring the electron yield atthe helicity switching frequency. The di�erence peaks at

both the L3 and L2 edges of Co are larger than those atthe Fe edges showing that the moment of the CoFeZr isconcentrated on Co. Scanning the photon energy acrossthe Zr L edges showed XAS features, but no evidence ofany polarization on the Zr.

0

1

2

3

4

5

690 720 750 780 810

CoFe

Energy (eV)

Ele

ctro

nY

ield

(arb

.uni

ts)

Figure 9: (Color Online) Electron yield as a function of energyacross the Fe (circles and line) and Co (line) L3 and L2 edgesfor the single 200 Å CoFeZr �lm. The absorption data is anaverage of two curves collected under opposite, but saturating�elds and has been normalized to give a step height of unityacross the two edges.-0.1

0

0.1

680 720 760 800 840

-2.0

-1.5

-1.0

-0.5

0

Energy (eV)

Ele

ctro

nY

ield

Diff

eren

ce(a

rb.u

nits

)

Cum

ulat

ive

Inte

gral

Figure 10: (Color Online) The di�erence in electron yield asa function of energy across the Fe and Co edges (circles andline) and the cumulative integral of the di�erence signal (line)for the single CoFeZr �lm. The data was recorded simultane-ously with the absorption signals shown in Fig. 9. Artifactsassociated with recording the electron yield in a �eld havebeen removed by averaging two curves collected under op-posite, but saturating �elds. No normalization or correctionterms have been applied to the data.The integral terms A and B were obtained by calcu-lating the cumulative integral of the di�erence data as afunction of energy across the two edges. From Eq. 1a,it is clear that the orbital moment is proportional to thesum of A and B. From the shape of the cumulative inte-gral in Fig. 10 it is clear that the orbital moment on the

7Fe sites is nearly quenched whilst there remains a con-siderable orbital component on the Co. As the elementspeci�c constant in Eqs. 1a and 1b was not known, it wasnot possible to extract the speci�c spin and orbital com-ponents, but the ratio is independent of C and can bedetermined easily. For Fe, the ratio of the orbital to spinmoments is 0.012± 0.005 and for Co, is 0.078± 0.005 forthe single CoFeZr layer sample. Thus, the contributionto the orbital moment is dominated by the Co contribu-tion.When the magnetic CoFeZr layer is in a multilayer thedata trends are remarkably similar. No di�erence be-tween the multilayer samples was observed and we showin Fig. 11 representative data for sample 30/30. Thedata is considerably noisier than that obtained for a sin-gle layer due to the increased depth of the magnetic layer.The detected electrons have to penetrate both the cap ofthe partially oxidized alumina and the top AlZr layer inthe multilayered structures. Again it is clear that the Feorbital moment is quenched. The ratio of the orbital tospin moments, in the multilayers are 0.01 ± 0.01 for theFe sites and 0.03 ± 0.01 for the Co.-0.10

-0.05

0

0.05

680 720 760 800 840

-1.5

-1.0

-0.5

0

Energy (eV)

Ele

ctro

nY

ield

Diff

eren

cex1

0(a

rb.u

nits

)

Cum

ulat

ive

Inte

gral

x10

Figure 11: (Color Online) The di�erence in electron yield asa function of energy across the Fe and Co edges (circles andline) and the cumulative integral of the di�erence signal (line)for sample (c). The data treatment is identical to that appliedto the data in Fig. 10. The lower signal-to-noise arises fromthe thickness of the e�ective capping layer which is the sumof the protective alumina layer and the top AlZr layer of thebilayer. IV. DISCUSSIONThe orbital contribution to the magnetic moment inamorphous CoFeZr layers is dominated by Co, as de-scribed above. The relative weight of the orbital contri-bution is larger in Co compared to Fe, in their elementalcomposition. The morbit/mspin were previously deter-mined using XMCD to be 0.043 and 0.095 for Fe and Corespectively17. Furthermore, the morbit/mspin is heav-ily a�ected by alloying; in Fe67Co33 these are 0.098 and

0.217 for Fe and Co respectively18. Thus the orbital mo-mentum of both Fe and Co is doubled by alloying, ascompared to their elemental values whereas the spin mo-ments are more modestly a�ected.In the CoFeZr alloys discussed here, the orbital con-tribution to the moment is suppressed for both Fe andCo. This is somewhat expected, as the ternary alloy isamorphous and therefore exhibits no long range crys-talline order. The resulting orbital to spin ratio of Cowas determined to be 0.078± 0.005 which is surprisinglyclose to what is obtained for pure crystalline Co (0.095).Thus, the absence of long-range order appears to havea minute in�uence on the orbit to spin ratio of the Comoment. The suppression of the orbit to spin ratio inFe is much larger, from 0.043 to 0.012± 0.005. Thus theamorphous structure suppress the orbital moment muchmore profoundly in Fe than Co when compared to theirelemental crystalline structures. The di�erences in thedegree of quenching can not be due to di�erences in thelocal con�guration of the atoms as they are random andexpected to be similar for both Fe and Co, as seen inthe similarity of the element speci�c magnetization loops(�g. 8). A strong contribution from the electronic partis therefore anticipated. Calculations of the in�uence ofboth the order and the alloying is therefore importantfor understanding these di�erences. There are currentlyno calculations available for these important material pa-rameters in amorphous alloys.The morbit/mspin contribution diminishes strongly forboth Fe and Co, when the CoFeZr layers are decreasedin thickness. This is apparent from the results obtainedon the CoFeZr/AlZr multilayers. The morbit to mspinratio is decreased to 0.01 ± 0.01 for the Fe atoms and0.03 ± 0.01 for the Co, in the 30/30 multilayer. Thus,there are strong proximity e�ects from the interfaces inthese materials. This is somewhat surprising consideringthe minute e�ect on the Co upon the formation of theCoFeZr alloy. Thus the contribution from the hybridis-ation of the ferromagnetic layer with the non-magneticlayers is likely to be important in this perspective. Sucha hypothesis is consistent with the microscopy and x-raydata in which intermixing at the interfaces is observed inall samples. However, a �rm conclusion depends on morecomplementary experiments or theoretical calculations ofthe interface e�ects.The Fe and Co are strongly coupled and can there-fore be treated as having magnetic properties which arethe weighted average of the two constituents. While theaddition of Fe increases the magnetic moment of the ma-terial, the orbital contribution is decreased. Thus, thecurrent measurements give a good understanding of thein�uence of Fe on the magnetic properties of amorphousCoFeZr alloys. The addition of Fe decreases the coerciv-ity, resulting in softening of the magnetic properties19,due to the decrease of the e�ective orbital contributionto the moment.In the amorphous alloys that we have presented, theabsence of crystalline order is apparently not su�cient

8to remove the orbital contribution to the moment. Thus,it is conceivable that local atomic con�gurations causesthe observed anisotropy in amorphous materials withthe subsequent spin orbit coupling being the underlyingcause of the magnetic anisotropy. Under normal growthconditions the con�gurational anisotropy is completelyrandom, due to the randomness of the atomic order.However, �eld induced anisotropy has been known forquite a while in Co rich magnetic amorphous materi-als20,21. Furthermore, the anisotropy can even be im-printed layer by layer in amorphous multilayers by theselection of the applied external �eld4. Thus the forma-tion of a macroscopic easy axis is possible, which must berelated to the local con�guration of the magnetic atoms.This con�gurational variation is most likely associatedwith small di�erences in the nearest neighbor distance.The length scales of these di�erences are out of reachfor direct structural measurements. The term con�gura-tional anisotropy would be a more appropriate descrip-tion of the obtained anisotropy in amorphous materials.V. CONCLUSIONThe morbit to mspin ratio of elemental Co is only mod-estly decreased by forming amorphous CoFeZr alloys.

The corresponding changes in Fe are much larger. Thus,the e�ective morbit to mspin ratio can be adjusted bythe Fe to Co ratio, allowing tailoring of the magneticanisotropy in amorphous CoFeZr alloys. Anisotropy cannot be treated as uniquely coupled to periodic latticesand the term crystalline anisotropy is therefore mislead-ing. The magnetic anisotropy in amorphous materialsmust be viewed as a local property, associated with theorbital contribution from atomic con�gurations and acorresponding local magneto-strictive �eld. Con�gura-tional anisotropy is therefore a more appropriate termwhen addressing the magnetic anisotropy in amorphousmaterials. The con�gurational anisotropy is a local prop-erty, which can even be aligned by applying external�elds during growth.AcknowledgmentsThe authors would like to acknowledge the supportfrom the Swedish Research Council (VR) and Knut andAlice Wallenberg Foundation (KAW).

∗ Electronic address: Bjorgvin.Hjorvarsson@fysik.uu.se1 D. Djayaprawira, K. Tsunekawa, M. Nagai, H. Maehara,S. Yamagata, N. Watanabe, S. Yuasa, Y. Suzuki, andK. Ando, Applied Physics Letters 86, 092502 (2005).2 J.-O. Song and S.-R. Lee, Journal of Magnetism and Mag-netic Materials 310, 1923 (2007).3 G. Suran, F. Machizaud, and M. Naili, Phys. Rev. B 47,15007 (1993).4 H. Raanaei, H. Nguyen, G. Andersson, H. Lidbaum, P. Ko-relis, K. Leifer, and B. Hjörvarsson (2009), accepted inJournal of Applied Physics.5 Y. Imry and S.-K. Ma, Phys. Rev. Lett. 35, 1399 (1975).6 R. Harris, M. Plischke, and M. J. Zuckermann, Phys. Rev.Lett. 31, 160 (1973).7 J. McCord, I. Monch, J. Fassbender, A. Gerber, andE. Quandt, Journal of Physics D: Applied Physics 42,055006 (6pp) (2009).8 A. T. Hindmarch, C. J. Kinane, M. MacKenzie, J. N.Chapman, M. Henini, D. Taylor, D. A. Arena, J. Dvorak,B. J. Hickey, and C. H. Marrows, Physical Review Letters100, 117201 (pages 4) (2008).9 P. Carra, B. T. Thole, M. Altarelli, and X. Wang, Phys.Rev. Lett. 70, 694 (1993).10 B. T. Thole, P. Carra, F. Sette, and G. van der Laan, Phys.Rev. Lett. 68, 1943 (1992).11 S. Pizzini, L. M. García, A. Fontaine, J. P. Rue�, J. Vogel,R. M. Galéra, J. B. Goedkoop, N. B. Brookes, G. Krill,and J. P. Kappler, Journal of Electron Spectroscopy andRelated Phenomena 86, 165 (1997).

12 J. Vogel, M. Sacchi, R. J. H. Kappert, J. C. Fuggle, J. B.Goedkoop, N. B. Brookes, G. van der Laan, and E. E.Marinero, Journal of Magnetism and Magnetic Materials150, 293 (1995).13 S. Imada, T. Muro, S. Suga, K. Kobayashi, H. Maruyama,and H. Yamazaki, Journal of Electron Spectroscopy andRelated Phenomena 78, 279 (1996).14 C. Bordel, S. Pizzini, J. Vogel, K. Mackay, J. Voiron, R. M.Galéra, A. Fontaine, P. Auric, J. B. Goedkoop, and N. B.Brookes, Phys. Rev. B 56, 8149 (1997).15 A. R. B. de Castro, G. B. Fraguas, P. T. Fonseca, R. N.Suave, S. Gama, A. A. Coelho, and I. A. Santos, Journal ofElectron Spectroscopy and Related Phenomena 101-103,725 (1999).16 M. Björck and G. Andersson, Journal of Applied Crystal-lography 40, 1174 (2007).17 C. T. Chen, Y. U. Idzerda, H.-J. Lin, N. V. Smith,G. Meigs, E. Chaban, G. H. Ho, E. Pellegrin, and F. Sette,Phys. Rev. Lett. 75, 152 (1995).18 L. Zong-Mu, X. Fa-Qiang, W. Li-Wu, W. Jie, Z. Jun-Fa,and Z. Wen-Hua, Chinese Physics Letters 24, 2667 (2007).19 G. Suran, M. Naili, H. Niedoba, F. Machizaud, O. Acher,and D. Pain, Journal of Magnetism and Magnetic Materi-als 192, 443 (1999), ISSN 0304-8853.20 G. Suran, K. Ounadjela, J. Sztern, and C. Sella, Journalof Applied Physics 55, 1757 (1984).21 G. Suran, J. Sztern, and B. Barbara, Applied Physics Let-ters 58, 1338 (1991).