Static and dynamic X-ray resonant magnetic scattering studies on magnetic domains - 6 AN X-RAY MAGNETO-OPTICAL STUDY OF MAGNETIC REVERSAL IN PERPENDICULAR EXCHANGE-COUPLED [PT/CO]n/FEMN MULTILAYERS

UvA-DARE is a service provided by the library of the University of Amsterdam (https://dare.uva.nl)

Static and dynamic X-ray resonant magnetic scattering studies on magnetic

domains

Soriano, J.M.

Publication date

2005

Link to publication

Citation for published version (APA):

Soriano, J. M. (2005). Static and dynamic X-ray resonant magnetic scattering studies on

magnetic domains.

General rights

It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulations

If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

6 6

A NN X-RAY MAGNETO-OPTICAL

STUDYY OF MAGNETIC REVERSAL

INN PERPENDICULAR

EXCHANGE-COUPLED D

[ P T / C O ]

/ F E M NN

MULTILAYERS

WeWe used X-ray magnetic circular dichroism and X-ray resonant magnetic scattering toto investigate the room-temperature exchange bias found in perpendicular anisotropy [Pt/Co][Pt/Co]nn multilayers coupled to antiferromagnetic FeMn films. About half a monolayer

ofFeofFe spins at the interface is found to be uncompensated. A fraction of these uncom-pensatedpensated spins are pinned in the exchange bias direction. The amount of pinned spins increasesincreases for smaller numbers of[Pt/Co] bilayers and seems to be responsible for the ex-changechange bias. This scenario, already observed at low temperatures for in-plane exchange biasbias systems, clearly applies also to the perpendicular counterparts at room temperature. RemarkableRemarkable differences are found in the magnetic correlation length for two samples and betweenbetween the forward and backward branches of the hysteresis loop.

6.1.. Introduction

Thee exchange-coupling effect observed in ferromagnetic /antiferromagnetic bilayerss has led in the last decade to important technological applications such ass spin valves and magnetic random access media. The effect was discovered

alreadyy in 1956 by Meiklejohn and Bean [156, 157], who observed an offset alongg the field axis of the hysteresis loop of the ferromagnetic layer and an en-hancementt of the coercivity which appeared only when the bilayer system was cooledd down through the Néel temperature under a saturating magnetic field. Althoughh already at that time it was clear that this effect was somehow con-nectedd to the F/AF interface, little progress in its understanding was achieved forr decades.

Severall models of the exchange-coupling phenomena have been pro-posed:: Meiklejohn assumed [158] a direct exchange coupling between the F and AFF spins at an ideal perfectly flat interface. The spins of the last AF monolayer att the interface, frozen into a certain direction by the neighbouring AF spins, wouldd couple to the adjacent F spins, breaking the symmetry of the F magnetic reversal.. This model results in an ad hoc unidirectional interface anisotropy that explainedd the hysteretic shift, but it implied much larger shifts than the obtained ones. .

Laterr it was realized that the realistic roughness at the F/AF interface wouldd reduce the number of F/AF spin pairs with direct interaction [159], thus supportingg the experimental data. However, perpendicular antiferromagnetic domainn walls near the interface [160,161] or spin-flopping in the AF layers [162] wouldd lead to the same result.

Anotherr key concept is the existence of uncompensated AF spins at the interface,, which will prefer to align with the F neighbours. Depending on whetherr these uncompensated AF spins flip with the F layer or not, they are saidd to be unpinned or pinned. Recently, element-specific X-ray spectromi-croscopyy allowed the direct observation of uncompensated antiferromagnetic spinss at the interface [163,164]. Other studies [165,166,167,168] have been de-votedd to the 3D structure of these spins. Furthermore, by using X-ray circular magneticc dichroism (XMCD), Ohldag et al. [169] showed that the interfacial AF unpinnedd spins increase the coercive field whereas the pinned spins produce thee exchange-bias field.

Althoughh almost all studies on exchange-coupled systems have focused onn systems with in-plane magnetization, very recently perpendicular exchange couplingg has also been found. Systems based on Pt/Co [109,166,167,170,171,

Ann X-ray magneto-optical study of magnetic reversal in perpendicular

exchange-coupledd [Pt/Co]n/FeMn multilayers 103

172,173]] and Pt/FeCo [174] multilayers with perpendicular magnetic anisotropy cann be exchange-coupled with several AF materials, such as Fe2F [170], CoO [109, 166,167,, 173], FeMn [171,174, 175] or NiO [172]. In the last two cases the ex-changee coupling persists well above room temperature. In general, the perpen-dicularr exchange bias is weaker, possibly as a result of the in-plane preferential orderingg of the AF compounds.

Inn this chapter, we present a magnetization reversal study of room tem-peraturee perpendicular exchange coupled F/AF films by means of XMCD and polarization-dependentt soft X-ray resonant magnetic scattering (XRMS). The motivationn for this study is to understand the mechanisms that govern the magnetizationn reversal as well as to study the role of the interfacial coupling strengthh on it.

Thee layout of the chapter is as follows: Sect. 6.2 gives an overview of the sampless and the current status of their understanding, Sect. 6.3 discusses the spectroscopic,, magnetization and scattering results and Sect. 6.4 summarizes thee conclusions.

6.2.. The [Pt/Co]

/FeMn perpendicular exchange bias

system m

Thee out-of-plane preferential axis of the Pt/Co multilayers originates fromm the Pt/Co interface anisotropics [176]. The perpendicular exchange-bias effectt induced by a FeMn overlayer was first discovered at the Commissariat a

VV Energie Atomique (CEA) facility, Grenoble (France) [174,177], which is also the

sourcee of our samples. An extensive review of the growth, structural and mag-neticc characterization of the system can be found in Ref. [175], which we will summarizee here.

Thee original samples were grown on Si substrates by magnetron sputter-ingg without purposely applying a magnetic field. However, a significant stray fieldd from the magnetron perpendicular to plane exists on the substrate during deposition.. The AF layers are therefore grown on Pt/Co multilayers that were nearlyy magnetically saturated. The samples were cooled under a perpendicular magneticc field from above 150 °C, which in turn gave the final exchange field

Pt t FeMn n Coo i Pt t n n Si3N„ „

Figuree 6.1: Cross section of the samples. Starting with a Pt layer, Pt/Co multilayer were depositedd 100 nm-thick SisISU membranes. The thickness of the FeMn layers was tAF =

100 ran. A Pt capping layer was added to prevent sample degradation. The number of bilayerss was 15 for sample A and 10 for sample B.

I?£.. Atomic force microscope measurements reveal a very low roughness (RMS << 1.5 A) and X-ray diffraction data showed a clear fee (111) texture in both the [Pt/Co]] multilayer and the FeMn layer.

Thee two-dimensional parameter space, spanned by the thickness of the ferromagneticc layer (represented by the number of bilayers n) and the AF layer thicknesss tAF, reveals a very rich phenomenology [175]. In the absence of the

AFF layer, the Pt/Co multilayer always presents perpendicular anisotropy, due too the interface anisotropy of the Co atoms, with a coercive field Be that in-creasess with the number of bilayers n [176]. Such free multilayers have a square hysteresiss loop as shown in the inset of Fig. 6.2, which also defines the coercive andd exchange bias fields.

Whenn such multilayers are capped with a 7 nm thick FeMn layer, the largee easy-plane anisotropy at the F/ AF interface causes an in-plane anisotropy forr n = 2. For n = 3, the competition between the in-plane and out-of-plane anisotropiess results in the formation of stripe domains at remanence, with a dif-ferentt exchange bias for the up- and down domains. For n > 4, the samples ex-hibitt out-of-plane anisotropy [178] and the hysteresis loops develop some tails relatedd to the inter-domain magnetostatic correlation [77].

Forr n > 4, both the coercive and exchange-bias field increase with the

ttAFAF== 10 nm

:: 0.4 nm 2.33 nm

Ann X-ray magneto-optical study of magnetic reversal in perpendicular

exchange-coupledd [Pt/Co]n/FeMn multilayers 105

Tablee 6.1: Magnetic properties of the [Pt/Co]„/FeMn multilayers.

Sample Sample A A B B n n 15 5 10 0 Be e (mT) (mT) 18.8 8 14.6 6 BE E (mT) (mT) 0 0 - 4 . 0 0

thicknesss t^F of the FeMn layer, the faster so for larger n. However, while the coercivee field soon saturates and decreases, the exchange bias field increases untill it saturates. Finally, for the thickest AF layers, both fields are equal and almostt independent of

t^F-Forr our experiments, we used a range of [Pt(2.3 nm)/Co(0.4 nm)]„ multi-layers,, where n is the number of bilayers, coated with a FeMn(10 nm) layer and aa protective 2 nm thick Pt layer. In order to allow transmission experiments, thesee samples were grown on 100 nm-thick Si3N4 membranes, which did not changee the properties. From this series, the results for two samples with n = 15 (samplee A) and 10 (sample B) bilayers are presented here. As can be seen from Tablee 6.1, only the latter sample shows exchange bias, but has a lower coercive field. .

6.3.. Results and discussion

6.3.1.. XMCD spectroscopy

Fig.. 6.2 shows the Co L23 normalized absorption and circular dichroic spectraa for both samples. The absorption spectra were measured in transmis-sionn in the flipping mode, in which the field is flipped at every energy point. In orderr to eliminate asymmetries caused by non-magnetic absorptive processes, thesee spectra were acquired for 1 light helicities and the resulting spectra were averaged.. The difference between the two absorption spectra, the XMCD sig-nal,, is plotted in the bottom panels of Fig. 6.2. For both samples, the asymmetry ratioo between the dichroic and the absorption R = ^+~(|_ amounts to 26%, in goodd agreement with XMCD measurements on C o / P d mutilayers [179].

Inn order to explore the magnetic state of the F/AF interface, we per-formedd Fe 1.2,3 XMCD measurements. Fig. 6.3 shows the absorption and dichroic spectraa for both samples. Despite the very poor signal-to-noise ratio, the

exis-A{n=-\5) exis-A{n=-\5) 6(n=10) )

11111111111 ii j1111111111ii

^^/^^

ii — i .

7700 775 780 785 790 795 800 805 810770 775 780 785 790 795 800 805 810

Photonn Energy (eV) Photon Energy (eV)

Figuree 6.2: Top panels: Co L23 absorption spectra of sample A (left) and B (right) ob-tainedd with parallel (dots) and antiparallel (line) alignment of photon spin and mag-netizationn vector. The spectra were scaled so that the intensity jump over the whole spectrumm is one. Bottom panels: circular dichroic spectra as the difference of the ab-sorptionn spectra. Inset: schematic of the relative absorption hysteresis loop, showing thee coercive and exchange fields Bc,£.

fencee of a very small circular-dichroic signal at the L3 edge indicates that some off the Fe spins are ferromagnetic and present a non-negligible projection in the out-to-planee direction. This is substantiated by the FeMn spin structure in the (111)) planes [180], that presents one spin oriented fully perpendicular to that plane.. At the F/AF interface, this uncompensated spin may flip along with the FF layer, giving rise to a circular dichroic signal.

Thee dichroic asymmetry ratio R, defined as the ratio of the maximum amplitudess of the L3 XMCD and helicity averaged absorption, is 1.7 0.6% forr sample A (n = 15), and 1.3 0.5% for sample B (n = 10). Using the bulk FeMnn lattice parameter of 3.63 A [168, 181] and a typical maximum dichroic asymmetryy ratio for fcc-Fe of R ~ 25% [182, 183], we find that these values correspondd to amounts of 1.8 0.7 monolayers (ML) of FeMn for sample A

Ann X-ray magneto-optical study of magnetic reversal in perpendicular

exchange-coupledd [Pt/Co]„/FeMn multilayers 107

A(n=:5) A(n=:5) B(hh =10) if) if)

a a

Photonn Energy (eV) Photon Energy (eV)

Figuree 6.3: Fe Y.2,3 X-ray absorption and circular dichroic spectra of sample A (left) and BB (right). As a guide to the eye, a B-spline function of the data of the bottom panels is shownn with a full line.

andd 1.4 0.6 ML for sample B. These values are somewhat larger than the ones obtainedd for in-plane systems by Ohldag et al. [169].

6.3.2.. Element-specific hysteresis loops

Inn order to verify the ferromagnetic behaviour of both layers, we mea-suredd the element specific hysteresis loops by measuring the field dependence off the dichroic absorption signal at the photon energy giving maximum con-trast.. The data shown in Fig. 6.4 are the result of averaging typically 10 loops, takenn at field sweep rates always lower than 10 mT/s. The data have been nor-malizedd to the helicity-averaged absorption pio = ^-/i{}i+ + }i~).

Inn the n = 15 sample A (left panel), the Co hysteresis loop presents a significantt coercive field of 18.8 mT and no exchange bias field, while the n = 100 sample B (right panel) shows a somewhat lower coercivity of 14.6 mT and a clearr exchange bias of -4 mT. Both loops reproduce the results measured with MOKE.. Since the coercive field of the free [Pt/Co] multilayer is 5 mT [175], it is

>A(n=15) )

B(nn =10)

-600 -40 -200 0 20 B(mT) B(mT)

400 60 -200 0 20

fl(mT) fl(mT)

Figuree 6.4: Element-specific perpendicular hysteresis loops taken at photon energies correspondingg to the maximum Co- and Fe-L3 XMCD. Full dots: Co data, left axis, openn circles: Fe data, right axis. Black and grey indicate respectively increasing and decreasingg fields.

clearr that the exchange coupling at the F/AF interface leads to an enhancement off the coercive field in both cases.

Withh respect to the Fe-L3 hysteresis loops (open symbols in Fig. 6.4), it iss clear that the uncompensated Fe spins follow the reversal of the F layer and thatt they are ferromagnetically coupled to the adjacent Co spins in both cases, whichh indicates that they are located at the F/AF interface.

Thee relative absorption loop of sample A shows the normal behaviour off a ferromagnetic system: it is symmetric with respect to the unpolarized ab-sorptionn }io, with a saturation amplitude of 1.6%. This symmetry indicates that alll the uncompensated Fe spins are unpinned, i.e., all flip when the F layer is reversed. .

Thee Fe hysteresis loop of sample B displays a reduction of the vertical amplitudee together with a vertical shift of the loop. Both effects can be under-stoodd in the frame of Ohldag's model [169], in which the presence of pinned spinss results in a gives a field independent XMCD amplitude and thus a ver-ticall shift of the XMCD hysteresis loop. Despite the poor quality of the data, especiallyy at the increasing field branch, we can estimate from this vertical shift thatt the number of unpinned uncompensated spins is about 1.3 ML, implying thatt the spins of 0.5 ML are pinned in a fixed direction.

Ann X-ray magneto-optical study of magnetic reversal in perpendicular

exchange-coupledd [Pt/Co]«/FeMn multilayers 109

6.3.3.. Scattering

Wee made use of the spatial resolution of XRMS to follow the magnetic domainn structure during reversal. In all cases, circularly symmetric scattering patternss where obtained. These were radially integrated to obtain the q depen-dentt scattered intensity I(qr)- These curves present a single broad peak,

com-parablee the data for GdFe discussed in Sect. 3.3.2.

Fig.. 6.5 compares the C0-L3 absorption hysteresis loops for both sam-ples,, shown in panels (a), with the data extracted from these curves. Panel (b) showss the average magnetic correlation length T, defined here as the distance overr which the magnetization is in the up or down direction obtained from the positionn of maximum intensity of the I(qr) curve. Panel (c) shows the width at

halff maximum £ of the diffraction peak, normalized to r, which is a measure of thee extent to which this correlation length is defined. Finally, panel (d) gives the totall scattered intensity.

Thee data for sample A (left) are symmetric in field. Since the intensity dataa for the positive field branch had less scatter, they have been used for the negativee branch also. The evolution of the magnetic correlation length in this samplee shows a quasi-parabolic shape: close to nucleation and saturation it di-vergess to values above 1 micron, corresponding to the small q limit set by the beamm stop. A minimum correlation length of about T = 440 nm is found in the fieldd range where the magnetization changes roughly linearly with the applied field.. As measured by the relative width of the scattering curve £ / T this corre-lationn length is also best defined in this range, although even at the maximum valuee the domains are completely uncorrelated in their relative positions.

Thee magnetic correlation length of the n = 10 sample B shows a markedly differentt behaviour, as shown in Fig. 6.5-c. The two branches of the hystere-siss loop for the opposite field directions display a strong asymmetry: while the branchh in the exchange-bias field direction (black data points) again has a quasi-parabolicc shape, the branch in the opposite direction (grey data points) shows largerr correlation lengths at nucleation, which fall off continuously to the sat-urationn field. This distinct behaviour is very likely due to the difference in the effectivee field acting on the ferromagnetic layer, which crudely speaking is the summ of applied and exchange-bias fields.

1.0 0 0.5 5 ££ 0.0 -0.5 5 -1.0 0 700 0 ~~ 600 ** 500 400 0 1.0 0 0.8 8 **** 0.6 0.4 4 1.0 0 0.8 8 22 0.6 ~"" 0.4 0.2 2 0.0 0 -600 -40 -20 0 20 40 60 -40 -20 0 20 40 66 (mT) B (mT)

Figuree 6.5: Absorption and XRMS data measured at the C0-L3 edge. Left: sample A (15 MLL Pt/Co), right: sample B (10 ML Pt/Co). Black/grey: increasing/decreasing field. Fromm top to bottom: a) hysteresis loops (full line); b) magnetic correlation length T derivedd from the maximum of the scattered intensity curve I(qr),c) ratio of the FWHM

off I(qr) to T, d) total integrated scattered intensity , compared with the 1 — (mz)2

curvee (line).

H o w e v e r ,, that this interpretation is too simple is clear from the fact that t h ee m i n i m u m d o m a i n size is attained at the s a m e absolute field v a l u e s in b o t h b r a n c h e s .. This s e e m s to i m p l y that t h e d o m a i n size is dictated m a i n l y by t h e a p p l i e dd field only, w h e r e t h e exchange field h a s the only role of shifting the n u -cleationn a n d s a t u r a t i o n fields. A(n-A(n-hh a / / == 15) //

-JJ \

JJ :

M M

--\\ J

K\ K\

-A A

Ann X-ray magneto-optical study of magnetic reversal in perpendicular

exchange-coupledd [Pt/Co]M/FeMn multilayers 111

Thee magnetic correlation length in both branches is even less defined thann in sample A. Since in thin films with perpendicular anisotropy the domain sizee is determined by the balance betweenn the gain in magnetostatic energy re-alizedd by the formation of domains versus the cost of creating domain walls, thiss difference is likely to be due to the stronger magnetization and therefore dipolarr interactions in sample A. However, the pinned uncompensated spins in samplee B may play a role by acting as an extra set of defects that hamper do-mainn wall propagation.

Magneticc reversal in systems with a high density of structural defects andd grain boundaries is usually explained by an activation energy of the Bark-hausenn volume Vg, which is the typical volume that is reversed in nucleation andd domain-wall propagation [184, 185, 186]. From time-resolved MOKE ex-perimentss [175] on samples with n = 4 and t&T = 7 nm, it has been inferred that Barkhausenn volumes Vg are smaller when the applied field is opposite to the exchangee bias. Although we can not identify the magnetic correlation length directlyy with the diameter of the Barkhausen volumes, the observed asymme-tryy in the former seems to support this picture.

Thee total scattered intensity shown in Fig. 6.5-d closely follows the 1 —

(m(mzz))22 curve in both samples. This is another confirmation of the applicability

off Parseval's theorem as discussed in Sect. 3.3.4, but also implies that the mag-netizationn is almost completely oriented perpendicular to the sample plane.

6.4.. Conclusions

Inn this chapter we investigated the origin of the perpendicular exchange biass that is found in ferromagnetic Pt/Co multilayers capped with an antifer-romagneticc FeMn layer. Specifically, two samples differing in the multilayer thicknesss were studied, where only the thinner system showed exchange bias.

Inn both samples, transmission XMCD measurements revealed the pres-encee of uncompensated Fe spins. Element-specific hysteresis loops showed that thesee uncompensated spins were directly coupled to the ferromagnetic mul-tilayer,, suggesting that they are located at the Co /FeMn interface. From these measurementss the amount of uncompensated spins was estimated as 1.8

mono-layerss of FeMn, which is about two times larger than found previously [169]. Forr the n = 15 sample that does not show exchange bias, the uncompensated spinss rotate with the F layer, i.e., they are unpinned. In contrast, in the n = 100 sample, about 15% of the uncompensated spins are pinned by the ferro-magneticc layer and do not reverse direction when the field is applied against thee exchange bias direction, apparently because the exchange interaction with thee ferromagnetic layer is weaker in this case. The existence of pinned uncom-pensatedd spins, which was already demonstrated in low-temperature in-plane exchangee bias systems, is therefore confirmed in this room-temperature perpen-dicularr exchange-bias systems.

XRMSS was used to measure the field dependence of the magnetic correla-tionn length. Both samples show highly uncorrected domain structures, typical forr a system with strong domain wall pinning by interfacial defects and grain boundaries.. Another consequence of this pinning is that the minimum corre-lationn length is not found at the coercive field as expected from micromagnetic theoryy (see Sect. 3.3.7).

Thee most notable result is the marked asymmetry in magnetic correla-tionn length for the two different field directions found in the n = 10 system thatt shows exchange bias. The fact that in both branches the minimum value iss attained at the same absolute value of the applied field suggests that the do-mainn evolution is determined by the applied field and that the exchange bias fieldd primarily shifts the nucleation and saturation fields. Clearly, the mag-neticc correlation lengths measured here are much too large to be interpreted directlyy as Barkhausen volumes. However, it is reasonable to suggest that the twoo are related, in which case our data would confirm the asymmetry in the Barkhausenn volume that was recently arrived at indirectly from time-resolved MOKEE measurements [187]. To make this point clearer, further transmission X-rayy microscopy experiments are required, as they would give more insight in thee detailed magnetic domain structure in these highly disordered samples.