O Japanese Society of Electron Microscopy Journal of Electron Microscopy 47(5): 379-385 (1998)

Review

Spin-polarized low energy electron microscopy offerromagnetic layers

T. Duden and E. Bauer

Department of Physics and Astronomy, Arizona State University, Tempe, AZ 85287-1504, USA.E-mail: bauer@asu.edu

Abstract Spin-polarized low energy electron microscopy (SPLEEM) is a powerfultechnique for the study of the magnetic domain structure of ferromagneticsurfaces, thin films and ferromagnetic / nonferromagnetic thin film systems.The unique feature of this instrument is the combination of magnetic withtopographic imaging on an atomic depth scale level, and with structuralinformation via low energy electron diffraction (LEED) from submicronregions. These capabilities are illustrated by studies of cobalt layers onW(110), Cu(l l l ) and Au(l l l ) surfaces, of Co-Cu and Co-Au bilayers and ofCo/Au/Co and Co/Cu/Co sandwiches.

Keywords spin-polarized low energy electron microscopy, ferromagnetic layersReceived 2 April 1998, accepted 26 June 1998

PrefaceA number of years ago, when LEEM was already wellestablished, Professor Moellenstaedt told one of theauthors that he had serious doubts initially that LEEMwould ever work but nevertheless recommended fundingof the further development of the instrument. This paperis dedicated to his memory and is a thank for his support.

IntroductionThin ferromagnetic films and ferromagnetic/nonferro-magnetic thin film systems such as double layers, sand-wiches and multilayers have attracted increasing attentionduring the last two decades because of their importancefor thin film magnetic memories and sensors and becauseof their rich variety of exciting physical phenomena.Over the years it has become increasingly clear that themagnetic properties of these films are extremely sensitiveto their microstructure. Microstructural studies are there-fore very important for understanding these properties.In the past, the most important method which allows thesimultaneous study of the microstructure and of themagnetic domain structure of films on surfaces has beenscanning electron microscopy with polarization analysis(SEMPA, for a review see ref. 1). More recently, magneticforce microscopy (MFM, for a review see ref. 2) has alsodeveloped into a powerful technique for this purpose.Both methods have the disadvantage that the surface is

scanned with a fine probe,which leads to long imageacquisition times. SEMPA, which uses secondary electronsfor imaging, in addition, is generally insensitive to surfaceroughness on the atomic depth scale. In this paper wedescribe another magnetic imaging method, SPLEEM (fora review see ref. 3) which does not have these short-comings, and we give some examples of its application.

Physical basis and instrumentationThe basis of the magnetic contrast in SPLEEM is theexchange interaction Vex = Zy J ^ - ij) Sj • Sj between theincident electrons with spin Sj and the target electronswith spin Sj at their positions rf and ij, J being theexchange coupling strength. In regions with preferredspin alignment Sj a nonzero magnetization M - Ij Sj existswhich causes a magnetic contribution - P • M to thescattered signal when the incident beam is spin-polarizedwith a polarization P ~ L; Sj. In crystalline materialsmagnetic contrast can also be understood in terms of thespin-dependent band structure (see Fig. 1) in which spinup and spin down electrons have different energies. Whenthe energy of the incident beam is below -1 eV in Fig. 1,the incident beam is reflected because there are no allowedstates in the crystal. Between about 1 and 2 eV thereflectivity is low for spin-up P and high for spin-downP for the same reasons. Above 2 eV there is still a magneticcontribution to the contrast due to the different densitiesof states in the two bands. In very thin magnetic films

380 JOURNAL OF ELECTRON MI C RO S C OP Y, Vol. 47, No. 5, 1998

Fig. 1 Band structure of Co above the vacuum level in the [0001]direction [3].

100

80

# 60

B 40

20

010

8

g 6S" 4

10 20E[eV]

30 40

Fig. 2 Specular asymmetry and reflectivity of a 6 ML thick Co film on[3].

the different k values of the two bands for a givenenergy E have an important consequence: the resonanceconditions for the standing waves which form in suchfilms—when the boundaries are sufficiently reflecting—differ somewhat. This causes pronounced oscillations inthe asymmetry A = (R+ -R_)/( R+ + R_) where R is thereflectivity of the surface and the + / - signs stand foropposite polarizations. This is clearly seen in Fig. 2 [4]which shows the asymmetry A (multiplied with thepolarization of the incident beam P » 25%) and thereflected intensity R of a 6 monolayer (ML) thick Co filmon W(110) as a function of energy. With increasingthickness the frequency of the quantum size oscillationsincreases rapidly so that they cannot be seen any longerin thick films (s»20 ML) because of the simultaneouspresence of several thickness levels. What remains are

Fig. 3 Spin-polarized low energy electron microscope (SPLEEM). Thepolarization manipulator is on the left-hand side.

the reflectivity oscillations due to the band structure ofthe crystal ('Bragg reflections, I(V) curve').

From what has been discussed up to now it is clearwhich features a SPLEEM instrument should have: (i) ahigh intensity In and polarization P of the incident beamfor optimum signal/noise ratio because of the usuallyweak magnetic contrast, (ii) the possibility to rotate P inany desired direction so that the magnetic contrast canbe optimized by orienting P parallel to M and (iii) thepossibility for rapid and flexible image accumulation andprocessing so that asymmetry images with good signal/noise ratio can be obtained rapidly. The first requirementis at least partially fulfilled by a Cs-O-activated GaAsphotoemission cathode, which emits a nearly monochro-matic intense beam of spin-polarized electrons whenilluminated with circular polarized red laser light, thesecond one by a spin-polarization manipulator consistingof a compound magnetic/electrostatic 90° deflector and amagnetic lens [5] and the third one by a suitable imageacquisition system. Otherwise a SPLEEM instrument islike a standard LEEM instrument with electrostatic con-denser and objective lens, the former in order to avoidrotation of the polarization when changing condenserexcitations, the latter in order to have no magnetic leakagefield at the specimen. The instrument used in the workreported here (Fig. 3) is completely surrounded byHelmholtz coils so that the films can be grown and studiedin the absence of magnetic fields.

Before presenting some results a few words are neces-sary about the nonmagnetic, that is topographic, contrast.Although the dominating contrast is usually diffractioncontrast which is determined mainly by the periodicity ofthe crystal normal to the surface, other contrast mechan-isms are more important in the present context: stepcontrast and quantum size contrast. Step contrast is dueto the destructive interference of waves reflected from

T. Duden and E. Bauer SPLEEM of ferromagnetic layers 381

JFig. 4 Step (a) and quantum size (b) contrast image formauon in LEEM. Field of view 6 \xm, electron energy 1.2 eV (a) and 3 eV (b). The sensitivitygradient of the detector from left to right has not been corrected. For explanation see text [6].

Fig. 5 Domain (c) domain boundary (d) image formation in in SPLEEM. Field of view 13 urn, electron energy 2 eV [7]. For explanation see text.

the two different height levels in the vicinity of atomicsteps and allows atomic roughness determination with alateral resolution of 10-20 run. Quantum size contrastallows detection of thickness variations on an atomicdepth scale with a lateral resolution of 20-50 run. Figure4 illustrates these two contrast mechanisms for a 5 MLthick Co film on a W( 110) surface consisting of 4, 5 and6 ML thick regions [6]. This high surface sensitivity doesnot necessarily mean low sampling depth. The inelasticmean free path of electrons rises very rapidly with decreas-ing energy below -10 eV so that back-scattering comesnot only from the topmost few layers but from manylayers, provided that the incident energy does not liewithin a bandgap. This can be successfully used to studythe correlations between the magnetization of top andbottom magnetic layers in sandwiches. Much work, how-ever, still has to be done do determine the influenceof spin-dependent inelastic scattering which in generaldecreases the magnetic contrast.

Figure 5 [7] illustrates magnetic contrast formation fora 6 Ml thick Co film on W(110) which has a strong in-plane anisotropy. Images (a) and (b) were taken with Pparallel and antiparallel to M in the domains with oppositemagnetization directions. They show weak magnetic con-

trast in addition to the step contrast of the substrate; (c)is the difference image of (a) and (b) which should showonly magnetic contrast. The remaining weak step contrastis due to a slight shift between the two images. Figure 5dis also a difference image but in this case the two Pdirections were perpendicular to M in the domains sothat only the domain boundaries are visible.

ResultsBefore the spin manipulator was available only the in-plane component of the magnetization could be measured.The strong magnetic contrast seen in thin Co layen onW(110) led to the conclusion that M was purely in-planewith a pronounced [1-10] easy axis. Later studies withthe polarization manipulator showed that this conclusionwas premature [8]. They revealed that there is a thicknessdependent out-of-plane component which varies on alength scale quite different from that of the in-planecomponent. Figure 6 illustrates this: (a) and (b) are in-plane and out-of-plane images, respectively; (c) is the tiltangle image; and (d) a topographic image taken at anenergy and focus at which substrate step and film quantumsize contrast are obtained simultaneously. The magnetiza-

382 J O U R N A L OF E L E C T R O N MI C RO S C O PY, Vol. 47, No. 5, 1998

Fig. 6 Magnetic images (in plane (a), out of plane (b) and tilt angle (c)) and topographic image (d) of a 5 ML thick Co film on VV(UO). Field ofview 8 Jim, energy 2.1 eV [8].

tion in this layer is wrinkled with the wrinkles occurringat the substrate steps independent of thickness fluctu-ations. The tilt angle, measured from in-plane, decreasesfrom -35° at 3 ML to -5° at 8 ML, indicating a strongpositive W/Co interface anisotropy. This occurs continu-ously, possibly slightly oscillatory without change of thedomain configuration.

A quite different dependence of the average magnetiza-tion direction and domain structure upon film thicknessis observed in the system Co/Au(ll l) . This system hasbeen studied extensively with SEMPA [9] and othertechniques, in particular its 'spin reorientation transition'(SRT), that is the transition from out-of-plane magnetiza-tion at small thickness to in-plane magnetization at largefilm thickness. The details of this transition, however, hadnot been clarified which stimulated its study with SPLEEM[10]. In order to see the details the Co film thickness wasincreased in the transition region in very small increments(0.05 ML) and images of all three magnitization compon-ents were recorded after each dosis. This is possible withSPLEEM because of the rapid image acquisition (8 s perasymmetry image). A few selected image triplets in thetransition range are shown in Fig. 7.

The study revealed that the SRT occurs in three steps:(i) Bloch wall broadening, (ii) magnetization switching ofsmall domains and (iii) Neel wall formation. The firstsigns of step (i) are already seen at 4.2 ML and dearlyevident at 4.3 ML. The broadening of the Bloch wallscreates in-plane magnetized regions closely related to theout-of-plane domain pattern. This fine-grained magnet-ization distribution is energetically unfavourable so thata small thickness increase is sufficient to induce step (ii)which already starts at 4.4 ML and is clearly evident at4.45 ML. At 4.5 ML the in-plane domains have grownsignificantly but there is still a noticable out-of-planecomponent causing a random fine-grained rippled mag-netization. To the extent to which the out-of-plane com-ponent disappears Neel walls develop between theincreasingly larger in-plane domains above 4.55 ML (seethe 4.7 ML images).

The reverse process, that is a transition from in-planeto out-of-plane magnetization can also occur, at leastpartially. Very thin metal overlayers are sufficient to

induce this transition which is attributed to the anisotropychange accompanying the replacement of the vacuum/Co interface by the metal/Co interface. Our old data [11]in which only the in-plane component could be measuredshowed that 1 ML of Au on a 10 ML thick Co film onW(110) had no influence on the in-plane magnetizationbut 2 ML turned M out-of-plane over most of the film,leaving only local in-plane domains which persisted withfurther Au deposition (Fig. 8). The origin of this anisotropychange, elastic or electronic, in particular its peakingaround 1 ML, is still a matter of debate. Electronic structurecalculations [12,13] for Cu and Au overlayers on Comonolayers on Cu(l l l ) and Au(l l l ) substrates show thesame trend as a function of overlayer thickness as theexperimental observations on thicker films (for referencessee refs. 12,13). However, no conclusions concerning thepossible contributions from elastic effects can be drawnfrom this qualitative agreement because of the Co thick-ness differences and because of the fact that these observa-tions were made with laterally averaging techniques.SPLEEM can measure the direction of M—which is deter-mined by the anisotropy—from regions with constantthickness via the quantum size contrast. As an exampleFig. 9a shows the topographic image of a 1.5 ML thickCu overlayer grown at 365 K on a Co layer. The M tiltangle obtained from the corresponding magnetic imagesat various Cu coverages are shown in Fig. 9b. The trianglesare from the uncovered regions, the full circles from the1 ML regions and the open squares from the 3 ML regionswhich result from the growth of 2 ML thick crystals ontop of the first ML. It is obvious that the perpendicularanisotropy is highest at 1 ML and that the double layercrystals growing on it sharply reduce the anisotropy inthe 1 ML regions. The different tilt angles at 1 ML Cu forML islands on the bare Co layer and ML regions between3 ML thick regions clearly show that overlayer-inducedanisotropy changes cannot simply be explained by themodified electronic structure of the first ML. Thicknessand micro structure dependent misfit strains seem to playan important role too.

Putting another Co layer on top of the nonmagneticoverlayer produces a sandwich. Sandwiches have beenthe subject of many studies because of the non-collinear

T. Duden and E. Bauer SPLEEM of ferromagnetic layers 383

4.4 ML 4.45 ML 4.5 ML 4.7 MLFig. 7 Different stages of the spin reorientation transition in a Co film on a 10 ML thick A u ( l l l ) layer on W( 110). 7X7 | im2 images, electron energy1.2 eV [10].

t - 1 ••.-,. v ;Jb. '*• , \ r

il- IFig. 8 Change of magnetization direction in a 10 ML thick Co film on W(110) (a) by a Au overlayer (b). Only the in-plane images with P in theeasy direction are shown. Field of view 13 urn, energy 1.5 eV [13].

(non-ferromagnetic) coupling between top and bottomlayer which occurs at certain interlayer thicknesses. Anti-ferromagnetic coupling between the two layers has beenmade responsible for the technologically important giantmagnetoresistance of multilayers and is usually deducedfrom hysteresis curve measurements. In previous oldstudies [7,11] we had observed at the Cu film thicknessof optimum antiferromagnetic coupling—according to thelaterally averaging measurements—pronounced non-collinear coupling with much smaller domains in the top

layer and M rotated in-plane over a wide angular rangearound 90". Our recent experiences with single and doublelayers have lead us to a more detailed study of thecoupling in Co/Au/Co and Co/Cu/Co sandwiches [14].As an example of the results Fig. 10 shows the couplingthrough a 6 ML thick Au layer which is just above thethickness of optimum antiferromagnetic coupling. The Aulayer attenuates the magnetic contrast somewhat but 3ML Co on top of it nearly completely suppress thein-plane contrast while a strong out-of-plane contrast

384 J O U R N A L OF E L E C T R O N MI C RO S C O P Y, Vbl. 47, No. 5, 1998

0.5 1 1.5 2Cu coverage [ML)

25

Fig. 9 Cu overlayers on a 5ML thick Co film on W( 110). (a) Topographic image at 1.5 ML thickness. E = 1.4 eV, field of view 9 urn (b) Thicknessdependence of the M tilt angle (measured from the in-plane direction) in regions of constant local thickness. For explanation see text [14].

6 Au/7 L u -i L o/f> All1 / 1- O 7 Co/6 Au/7 Co

Fig. 10 Relationship between top and bottom layer magnetization in a n ML Co/6 ML Au/7 ML Co sandwich on W(l 10). 9X9 nm2 images, electronenergy 1.4 eV [14].

develops, just as in the case of the growth on the Au (111)surface. The out-of-plane domain structure perfectly rep-licates the weak out-of-plane pattern of the Co underlayer.With increasing Co top layer thickness the transition toin-plane magnetization occurs again similar to that seenin Fig. 7 on Au(lll) but now SPLEEM shows twoapparently unrelated in-plane images: one perfectly replic-ating the domain pattern of the substrate, the other with90° rotated M direction which is typical for biquadraticcoupling, usually attributed to interface roughness. Vectoraddition of the two M components leads to a wrinkledin-plane magnetization in the top layer.

ConclusionThe examples presented here give only a hint of theinformation on the magnetic microstructure and its cor-relation with the crystalline microstructure which can beobtained with SPLEEM. There are still many problemsinvolving single, double and triple layers awaiting investi-gation at constant temperature and zero magnetic field.Changes of the correlations between the magnetic andcrystalline microstructure with temperature have alreadybeen studied in one example [15]. Real time studies inmagnetic fields normal to the surface are possible buthave not been made up to now. Improvements in image

T. Duden and E. Bauer SPLEEM of ferromagnetic layers 385

acquisition rate and resolution also still can be expected.Subsecond imaging with less than 10 nm resolution andsufficient contrast at a good signal/noise ratio appearfeasible. This will significantly increase the possibilities ofSPLEEM.

AcknowledgementsThe studies reported here were supported by the Deutsche Forschungsge-meinschaft and Arizona State University. The loan of the SPLEEMequipment by the Technische Universitaet Clausthal is appreciated.

References1 Unguris J, KelleyM H, Gavrin A, Celotta R J, Pierce D T, and Scheinfein

M R (1997) Scanning electron microscopy with polarization analysis(SEMPA). In: Handbook of Microscopy, Methods U, eds Amelinckx S, vanDyck D, van Landuyt J, and van Tendeloo G, p. 735, (VCH Weinheim).

2 Wadas A (1997) Magnetic force microscopy. In: Handbook of Microscopy,Methods II, eds Amelinckx S, van Dyck D, van Landuyt J, andvan Tendeloo G, p. 845 (VCH Weinheim).

3 Bauer E (1997) Spin-polarized low-energy electron microscopy. In:Handbook of Microscopy, Methods II, eds Amelinckx S, van Dyck D,van Landuyt J, and van Tendeloo G, p. 751 (VCH Weinheirn).

4 Wurm K (1994) Spin-polarisierte LEEM-Untersuchungen an duennenKobalt-Epitaxieschichten auf W<110), MS thesis, TU Clausthal.

5 Duden T and Bauer E (1995) A compact electron spin polarizationmanipulator, Rev. Sd. Instrum, 66: 2861.

6 Duden T (1996) Enrwicklung und Anwendung der Polarisations-manipulation in der Niederenergie-Elektronmikroskopie, PhD Thesis,TU Clausthal.

7 Bauer E, Duden T, Pinkvos H, Poppa H, and Wurm K (1996) LEEMstudies of the microstructure and magnetic domain structure ofultrathin films. J. Magn. Magn. Mater. 156: 1.

8 Duden T and Bauer E (1996) Magnetization wrinkle in ultrathinmagnetic films, Phys. Rev. Lett. 77: 2308.

9 Oepen M, Speckmann M, Millev Y, and Kirschner J (1997) Unifiedapproach to thickness-driven magnetic reorientation transitions, Phys.Rev. B 55: 2752 and references therein.

10 Duden T and Bauer E (1997) Magnetic domain structure and spinreorientation transition in the system Co/Au/W(110), MRS Sytnp.Proc. 475: 283.

11 Pinkvos H, Altman M, and Bauer E (1993) unpublished results.

12 Zhong L, Kim M, Wang X, and Freeman A J (1996) Overlayer-inducedanomalous Interface magnetocrystalline anisotropy in ultrathin Cofilms, Phys Rev. B 53: 9779.

13 Ujfalussy B, Szunyogh L, Bruno P, and Weinberger P (1996) First-prindples calculations of the anomalous perpendicular anisotropy ina Co monolayer on Au(l l l ) , Phys. Rev. Lett. 77: 1805.

14 Duden T and Bauer E, to be published.

15 Pinkvos H, Poppa H, Bauer E, and Kim G-M (1993) Time-resolvedSPLEEM study of magnetic microstructure in ultrathin Co films onW(110). In Magnetism and Structure in Systems of Reduced Dimension,eds Farrow RFC, Dieny B, Donath M, Fert A, and Hermsmeier B D,p. 25 (Plenum, New York).