Magnetism and structure in hexagonal Fe/Ru superlattices with short periodicity

Journal of Magnetism and Magnetic Materials 93 (1991) 15-24 15 North-Holland

Magnetism and structure in hexagonal Fe/Ru superlattices with short periodicity

M. Maurer "" ~, M. Piecuch a, M.F. Ravet% J.C. Ousset% J.P. Sfinchez b, C. Aaron b, J. Dekoster c, D. Raoux d, A. De Andres <e, M. De Santis ~, A. Fontaine d, F. Baudelet d, J.L. Rouvi~re b and B. Dieny f

~Laboratoire Mixte CNRS Saint Gobain, CR Pum, B.P. 109, 54704 Pont-h-Mousson, France bCENG / DRF, Sercice de Physique, B.P. 85 X, 38041 Grenoble, France Clnstituut coor Kern- en Stralingsfy'sika, K.U. Leucen, 3030 Leucen, Belgium aLURE, B~t. 209(t, 91405 Orsay, France ~Departamento Ffsica Aplicada C IV, Unicersidad Auton6mica de Madrid, 28049 Madrid, Spain fLaboratoire Louis N~el, (WRS, B.P. 166 X, 38042 Grenoble, France

We present recent results on the structure and on the magnetic properties of Fe /Ru superlanices. Detailed knowledge of the local structure is obtained from X-ray diffraction, including in the anomalous mode, and by EXAFS. thus bringing evidence that Fe layers correspond to an expansion exceeding 10%. The magnetic state is mostly characterized by M6ssbauer spectroscopy and also by magnetization measurements, thus demonstrating that the internal Fe layer carry a moment of about 2~B, whereas the two interfacial layers bear no moment. Except for the extension of the non-magnetic interface, these results are in agreement with theoretical predictions of Kiibler. which indicate an high-spin metastable state for such an expanded hcp Fe phase.

I. Introduction

Materials which are artificially kept far from their equilibrium state often display very in- triguing properties. In particular, there is a long standing interest in exploring magnetic transition metals in crystalline forms which are not the stable ones. With this view, one of the most exciting theoretical prediction is the existence of a so-called high-spin state for most transition metals at large volume expansion [1-5]. Epitaxy of pseudomorphous layers is basically the only manner to obtain controlled layers of elemental metals with artificial variation of the structure and of the interatomic distances. In this context, the case of metallic Fe is worth of attention. Indeed, besides the bcc and fcc structures, an hcp

~Present address: Compagnie de Saint Gobain, Direction de la Recherche, Les Miroirs, 92096 Paris-La D~fense Cedex 27, France.

phase is obtained under high pressure [6]. Only the bcc (8-coordinated) one is magnetic, whereas the fcc and hcp (12-coordinated) forms, which

o 3 correspond to a volume of l l . 4 A / F e atom, are non-magnetic. According to theoretical predictions, both fcc and hcp bulk phases should un- dergo a transition to a magnetic state at volume expansions of the order of 5% [5, 7]. The purpose of this paper is to discuss our present knowledge of the structure of (0001) hcp Fe /Ru superlattices [8] and to elucidate the effect of the Fe layers stretching on the magnetic properties. Since magnetic properties are extremely dependent on the atomic scale environment, we have thor- oughly investigated the local structure around Fe atoms, including the interatomic distances, the nature and the number of near neighbours. Re- garding magnetism in particular, it is essential to characterize properly atomic roughness at the Fe /Ru interface layer because Ru is very effi- cient to destroy Fe magnetism in Fe,Ru~_, hcp

0304-8853/91/$03.50 © 1991- Elsevier Science Publishers B.V. (North-Holland)

16 M. Maurer et al. / Hexagonal Fe / Ru superlattices with short periodicity

solid solutions [6, 9]. Indeed these alloys remain non-magnetic up to their limit of stability, i.e. around x = 75at% [10], although Fe atoms are surrounded by 8 Fc nearest neighbours on average. This effect is resulting from a strong hybridization of the 3d Fe bands with the 4d bands of Ru.

In the present report, we focus on the structural investigations which combine high-resolution and anomalous X-ray diffraction and also EXAFS. In the latter experiments, the use of polarized X-ray beam is fruitful, since angular dependence can be analyzed. In a next section, we focus on local magnetic properties, which are investigated by 57Fe M6ssbauer spectroscopy in

transmission geometry [11]. The link with bulk magnetization will also be discussed. In the last section, we describe preliminary results obtained with a new series of F e / R u superlattices which where grown on a Ru buffer layer, oriented (112,2) [12]. Rather surprisingly, atomically flat surfaces are obtained in spite of the non-dense nature of the (117,2) planes and a pseudomorphous growth of Fe has further been discovered, opening the possibility to elaborate a new class of superlattices where the growth direction breaks the origi- nal 6-fold symmetry. This dramatic lowering in crystal symmetries is thought to lead to superlattices with a monoclinic crystal structure, strictly speaking. Magnetic properties of these latter materials are yet to be elucidated.

2. Experimental procedure

The (0001) F e / R u superlattices were either grown on (117.0) sapphire substrates or on freshly cleaved mica discs in a RIBER EVA 32 MBE. In every case, a Ru buffer layer of 200.~ was firstly grown at 500°C to ensure an atomically flat and single crystalline surface. For (117,2) Ru growth, it was necessary to prepare thicker buffer layers (400 to 600.~) on (1102) sapphire before a good Ru surface was obtained [8, 12]. Superlattices were then elaborated at 100°C. The working pres-

sure never exceeded 5 × 10-10 Torr during deposition. X-ray diffraction data were obtained with a high-resolution goniometer using a 4-crystals monochromator (Co K 1). The anomalous diffraction experiments were carried out at LURE with a 2-circles goniometer using a Si(200) 2- crystals monochromator. Anomalous effects were achieved by recording spectra l(k, E,1) at different energies (e.g. E n = 6539, 7112, 7709, 8333 or 15203eV) surrounding the Fe K edge [13]. Owing to the small thickness of the samples, EXAFS spectra were recorded by counting all secondary electrons [14]. Great care was paid to the determination of a relevant set of phaseshifts and back-scattering amplitudes, by adjusting them in Fe ,Ru I , h c p solid solutions [10, 15]. M6ssbauer spectroscopy experiments were carried out with a stacking of layers from the same batch, in a standard transmission geometry [11]. Magnetiza- tion measurements were performed with a SQUID at temperatures ranging from 4.2 to 300 K.

3. Structure of the (0001) Fe /Ru superlattices by diffraction

The RH EED diffraction patterns [8, 12] of the Fe layers grown on a flat and single crystalline Ru surface, ascertain that Fe planes stack pseudo- morphically on the (0001) dense planes of Ru. Around 6 or 7 Fe monolayers, the diffuse scattering increases significantly whereas the diffracted streaks slightly deteriorate, as a consequence of modification in the structure. In spite of this change, we have once attempted to deposit thicker Fe layers, i.e. up to 10 monolayers (21A). Then, the R H E E D patterns reflect some disorder and roughening, but it is remarkable that no further degradation occurs upon stacking the next 42,~ Ru layers, while the symmetry of the plane clearly remains hexagonal. Such modifications certainly correspond to a large density of epitaxy disloca- tions. As a matter of fact, X-ray diffraction does not reveal any dramatic structural modification in

M. Maurer et al. / Hexagonal Fe / Ru superlattices with short periodicity 17

such a Fe(21,~)/Ru(42.&) superlattice, as compared to those with Fe thinner than 12,~.

The crystalline structure of these superlattices has formerly been discussed, with a particular emphasis on the nature of the stacking (ABAB. . . or A B C A B C . . , ) [8, 12]. At present, the most convincing arguments in favor of the hcp structure merge from powder diffraction experiments and from the observation of (1123) diffraction peaks. Considering the basic hcp unit cell of (0001) F%/Ru~. superlattices (x and y represent

o

the Fe and Ru thicknesses in A, respectively), we observe a large variation of the mean parameter c(x, y). The intrinsic CF~ parameter for an hypo- thetically pure Fe layer can be approached by extrapolating the variations of c(x,y) towards pure Fe, i.e. to y = 0. This leads to a value of

o

Cve = 4.15A, which is surprisingly large. Indeed, we must keep in mind that the Fe basal plane is stretched by the pseudomorphism with Ru, with an interatomic distance of about 2.71A. The

o

enormous mismatch between aau(2.706A) and typical F e - F e distances in compact planes (2.50,~) induces strains and possibly, in-plane dis- locations. Indeed, we observe a small in-plane relaxation of 2.5%, when extrapolating again to a pure Fe layer [8]. Whether perfect pseudomorphism or full relaxation is assumed, the specific volume for pure Fe layers reaches either 13.15 or 12.6A3/Fe atom. The first value is rather relevant for thin Fe layers sandwiched by thick Ru layers in dislocation free regions whereas the second value would rather reflect the minimum average volume in fully strained superlattices. As a consequence, we must underline that the hcp Fe structure corresponds to a value of c/a between 1.537 and 1.572. By reference to the equilibrium volume ( l l . 4 A 3) in the fcc 7-Fe or in the hcp e-Fe phase (after correction for pressure), the atomic volume is expanded by 10% at least, i.e. far more than the critical volume for the transition to a high-spin state. This unusually large expansion raises the question whether this new stretched hcp structure could correspond to a metastable magnetic state, as strongly suggested

by the recent calculation of Kfibler [7]. Indeed, the stability up to 7 monolayers is surprising. Even though the interfacial F e / R u chemical in- teraction is guessed to be rather strong [16-18], these chemical forces do probably not extend beyond 2-3 monolayers. In the absence of any metastable energy minimum, they would likely be too weak to counteract the enormous cumulative elastic energy which is contained in Fe layers as thick as 7 monolayers [18]. This point will be discussed below when considering monoclinic (112,2) superlattices.

The characterization of the sharpness of the modulation gradient is particularly crucial in magnetic superlattices, since Fe moments are dramatically dependent on the number of Ru atoms in the neighbouring shells. Thus, the key point is the determination of the interdiffusion at the level of 1 monolayer at the F e / R u interfaces. Auger electron spectroscopy studies did not reveal any sign for diffusion of Fe into Ru. Indeed, with the typical electron mean free path of about 5,~ which corresponds to the Fe LMM and Ru MNN Auger signals, it is difficult to detect interdiffusion at the monolayer scale and it would not reflect the interdiffusion state of the whole superlattice. Attempts to detect kinks in the variations of Auger signals [19-21] have failed here, mostly because the deposition rate from e-guns cannot be controlled sharply enough. However, we also have studied the growth of Mn on the (0001) Ru surface in the same conditions, except for using effusion cells as Mn sources. In this latter case, kinks are observed at least for the two first Mn monolayers, reflecting that our Ru surface is suit- able for a layer-by-layer growth of transition metals without volume interdiffusion. An accurate insight into interface sharpness can be gained from diffraction spectra, and more precisely by anomalous diffraction experiments [13]. Measure- ments have been carried out on two samples, one with x = 12A and y = 16A, and one with x =

o

y = 4A, the latter one being particularly significant with respect to the interface structure and to the key point of the commensurability of the

18 M. Maurer et al. /Hexa~,onal Fe / Ru superlattices with short periodiciO'

superlattice period with the unit cell parameter c(x, y).

A first set of information is obtained by careful simulation of the superlattice with a nominal value of the modulation length of ,,1 = 28A i.e. x = 1 2 A and y = 16A. The real value of A is

9 ° 2 .6A. Variations of the diffraction peak intensity reveal that the fluctuations in the number of planes of each species do not exceed + 1 around the mean values of n(Fe) = 6 and n(Ru) = 8. The fluctuations are the smallest possible, keeping in mind that the stop-signal during deposition is given arbitrarily when a nominal thickness is

reached, independent of the fact that a metal layer is complete or not. The anomalous effect at the satellites of the (000 2l) [for l = 0, 1,2] Bragg peaks are very sensitive to the difference 8d between dFe_ w and dnu nu, i.e. the F e - F e interplanar distances and the R u - R u interplanar distances. This difference 8d between the two interplanar distances can be considered as a position parameter in the supercell of parameter A. Data reduction unambiguously points to ~d =

o

0.11 A, i.e. a value larger than the amplitude of o

the variations of 8c(x, y)/2 = 0.07A from x = 0 to y = 0. This result is a priori surprising because

it corresponds to an hcp local unit cell parameter in Ru of c = 4.33,~, (as compared to 4.28A in pure Ru)and to c = 4.11 ,~ in Fe (as compared to the above extrapolation of c=4.15,~,) . In fact, the two sets of data can easily be reconciled by considering that the hcp unit cell parameter c(x, y) merely reflects the distances in a virtual crystal. The fact that Fe planes, in the real structure, are slightly closer than evaluated above is not surprising. Nevertheless, the real volume per Fe amounts to 12.9,~ ~ per atom, by taking the true value of the basal plane parameter a = 2.69 A, measured in transmission geometry. Identical

o o

values are also found in the F e ( 4 A ) / R u ( 4 A ) superlattice. The enhancement of 0.035 A, for Ru interplanar distances is paradoxical at first sight. However, we must remind that, in a fully strained superlattices, the basal plane is relaxed, corre- sponding to a reduction of the specific surface

per Ru atom of the order of 1 to 2%. This likely results in a shear, with an expansion along the out-of-plane direction of the order of 1%. The second order strain along the c-axis, is directly deduced from the broadening of several order of Bragg peaks belonging to the (000 21) family. Its magnitude amounts to Ac/c = 0.3%, which is at least one order of magnitude smaller than the relative difference between Fe-Fe and R u - R u interplanar spacings (5.7 or 3.3%) according to the above discussion. This leads to the conclusion that the number of atomic steps, where Fe planes are connected to Ru planes, is negligible, thus

implying small interfacial roughness. A great wealth of information on interface is

o

further obtained with a superlattice F e ( 4 A ) / o

Ru(4A), i.e. where about 2 Fe planes alternate with 2 Ru planes. Despite all the atomic planes correspond to interfacial ones, a set of satellites is unambiguously observed at low and high angles (fig. 1). This is a direct evidence that the interface composition is sharp at the monolayer scale. In fact, the superlattice period A is so small that its commensurability with the unit cell parameter must be questioned. This means that, e.g. the satellite located around k = 2.2A ~ can be as well considered as the 3rd order satellite to the (0000) peak and as the - 1 order to the main (0002) Bragg peak. Considering the shape of these satellites with more attention, one observes that in each of them, there is a sharp "subsatellite", marked by an arrow in the fig. 1. Examination of the subsatellites position reveals that they are perfectly commensurate with the set of the (000 2/) Bragg peaks of the hcp unit cell. This is a direct proof that a part of the superlattice has a superlattice period A which is exactly commensurate with the unit cell parameter c(4,4) i.e. A = 2 × c(4,4) = 8.42A,. Another implication of this approximate commensurability concerns the intensity of all the even satellites, whose intensity practically vanishes (fig. 1). In terms of the atomic arrangement, such a result demonstrates that a large part of the superlattice is built from a perfect stacking of 2 Fe + 2 Ru planes, coherent


z

12. §0

7.45

~. 40 ..

O.

i n i . , a '1 u

Fe(4 .~)/Ru(4 .~,) /=,

Ic°mme"su'ate I I "subsatellites" ]

0002)

I I

1. 2. 3. 4. 5.

ee

k(,~-l)

( 0 0 0 4 )

6. 7 .

°~

,t:

R

F e

,q

)Fe~ ,'

l I ( I F E=7709~ - (FF E~7112)

Ru Ru ____ IF RU4/FE4 E : 7 1 1 2

i! I Ru i Fe Fe /! i ii /!

! i / /I Vv~

q ,

I I 0. 10. 20. 30.

R ( ~ )

(d is tance f rom the cen te r of t h e F e layer)

Fig. 1. Top: diffracted intensity of the hcp (0001) Fe(4~,) /Ru(4A.) superlattice. The (000 21) Bragg peak of the hcp structure is indicated, as well as the satellites due to the superlattice period A. The arrows indicate the set of subsatellites which are exactly commensura te (A = 2c(4,4) with the hcp unit cell. Bottom: (a) Fourier transform of the diffracted amplitude A(k, E) deduced from the top curve; (b) Fourier transform of the difference A(k, E = 7709) - A ( k , E = 7112). This picture directly represents the density of Fe along the c axis, with the origin at the center of a Fe layer. Notice the coherence between the Fe plane up to about 10 atomic planes. The calculation is unavoidably disturbed by cut-off in the Fourier transform, due to finite available k-range [22].

20 M. Maurer el al. / ttexagonal Fe / Ru steperlattices with ~hort periodiciO'

over a length large enough to give rise to a diffraction pattern. Obviously, the residual dc- fects and the imperfection in the deposited thickness contribute as an incommensurate set of

satellites, which are significantly broader and ar-

bitrarily shifted as compared to the commensurate set. Notice e.g. that the first incommensurate satellite is at a k-value larger than the first commensurate one whereas the situation is reverse for the second satellites. From these small dif- ferences, we can further conclude that the average " incommensura te" modulation is slightly

o

shorter than 8.42A, certainly because the nominal modulation was only of 8 A.

Simulation of such an ultrashort period superlattice would be very exciting. In fact, reaching a good agreement is particularly difficult because many effects have to be taken into account: first the superlattice period fluctuations (as discussed above), second, the possible substitution of Fe by Ru and vice-versa in each plane, third the real interplanar distances between Fe-Fe , F e - R u and R u - R u planes, fourth the strains, fifth the finite crystallite size and, last but not least, all the correlations which can exist between these phe- nomena. Too many assumptions should be done to unravel such a structure. Then, a completely different procedure has been employed, which is particularly adapted to the anomalous diffraction data [13]. The trick is to inverse the diffracted amplitude A(k, E) = _+~/(l(k, E)) data by Fourier transformation back to the real space as ex- plained below. Then, once the origin is given, like e.g. at the center of a Fe layer for obvious symmetry reasons, one has to take the square root of l(k, E) and thus to assign the + or - sign to the various peaks. Indeed, making the assumption that no amplitude nods fall within every peak (including satellite), one can easily estimate if the sign is + or - by starting from a perfect superlattice built from a stacking of 2 Fe + 2 Ru planes. It is easy to demonstrate that only major structural changes would invert the sign of the diffusion factor. Then, the amplitude spectra are Fourier transformed to the real space, thus pro-

viding directly the density of scattering centers versus the distance R from the origin, i.e. here the center of the Fe layers. The major advantage of this procedure is that very few assumptions arc made and that all the relevant information which is contained in the spectra, is taken into account for the determination of the structure (fig. 1). Consider now the energy dependence of A(k, E); around the Fe K-edge (7111eV), the complex scattering factor of Fe strongly changes whereas no significant change occurs with the Ru contri- bu t ion . T h e n , the a m p l i t u d e d i f fe rence 8A(k, E) /SE directly reflects the Fe "sublatt icc" structure since all the Ru contributions anhilate. By making again the Fourier transform to the real space, one directly obtains the correlations between Fe planes (fig. 1). The preliminary conclusions which merge from the above analysis arc the following: first, the structural coherence along the growth direction compares with the thickness of the superlattice, meaning that layers are basically single grains along the c-axis, second, the F e - F e and R u - R u interplanar distances arc perfectly consistent with the above values, third, the composition modulation is sharp at short distance and is nicely coherent over about 10 atomic planes (statistically over the entire sample) as a proof of a good layer-by-layer growth. Details on the whole procedure will be given elsewhere [22]. At this stage, it is worthwhile considering microscopic structural probes, including EXAFS in particular.

4. Local structure investigated by EXAFS

Many (0001) F e , / R u , , superlattices have been investigated by EXAFS [23] and only the major results are reported here. When the linear polarization of the X-ray beam is in the sample plane, the weight from the in-plane neighbours is in- creased as a consequence of the dipolar selection rules. In all the superlattices under investigation, a decomposition in 2 subshells is strongly suggested by a beat in the ~¢(k) curves. Then two F e - F e distances are deduced, one for the in-


plane neighbours [(2.73 _+ 0.03),~] and one for the out-of-plane neighbours [(2.52_+ 0.03)A], to be compared with the distances deduced by diffraction i.e. (2.68_+ 0.03) and (2.58 + 0.02) .~ re-

spectively. Regarding the nature of the nearest neighbours to Fe, the number of Ru atoms in the plane is always negligibly small, except possibly in

o o

the F e ( 4 A ) / R u ( 4 A ) where a maximum of 10% substitution is detected. This is again a local proof that intermixing is negligible. Besides EX- AFS, which is only sensitive to pair distributions (single scattering approximation), the shape of the absorption edge (XANES) is a powerfull probe for multiple correlations, i.e. for the symmetry of the atomic structure. Since XANES calculations are intricate, a simple comparison of the Fe XANES in F e / R u superlattices is performed with the XANES of transition metals in stable phases, including 12-coordinated fcc-Ni or hcp-Co or bcc-Fe [23]. A striking observation is that, when the polarization is in the basal plane, the Fe XANES is identical to the XANES of 12-coordinated Ni or Co, as a clear evidence for an hcp local structure. When the polarization is perpendicular to the basal plane, the XANES departs from a typical XANES for hcp metals and shows some similarity with a bcc XANES. However, we must keep in mind that the hcp Fe planes have a very anomalous c/a value, i.e. a strong distortion of the nearest neighbours shell, which could modify the XANES. This point is yet to be elucidated.

To conclude this first part where the structure of (0001) hcp F e / R u superlattices has been discussed, we can state that Fe layers are stabilized in an hexagonal structure, with an anomalous volume of (12.8 _+ 0.2),~3/Fe atom, correspond- ing to an expansion of more than 10% as compared to the y-Fe or e-Fe phases. A full set of experiments brings the unambiguous conclusion that the interfacial intermixing remains extremely small but that most of the interfaces contains some steps. Some defects at larger scale including tilts and twins, are certainly present, as recently

observed by cross-sectional HRTEM but, again, sharp interfaces are observed [24].

5. Magnetic properties investigated by M6ssbauer spectroscopy

57Fe M6ssbauer spectroscopy experiments have been performed with the aim to gain a microscopic view on the Fe magnetism and also to characterize the local structure around Fe atoms. A first major result concerns the local symmetry, which is approached through the electric field gradient tensor. Indeed, all the sample (except the superlattice with a large Fe thickness x = 21 A, y = 42,~) exhibit a doublet at room temperature, with a well-defined quadrupolar interac- tion resulting in an asymmetric split spectrum (fig. 2). This quadrupole splitting is clearly origi- nating from the local axial symmetry (6-fold). Indeed, a rotation of 0 = 54 ° of the sample c-axis versus the direction of propagation of y-rays results in a variation of the relative intensity of the two transitions [11]. Although these variations reflect some imperfections in the crystal symmetry and in the c-axis orientation, they are a clear- cut proof for an hexagonal local structure at the Fe sites. The magnitude of the quadrupole splitting A slightly depends on the Ru thickness in the superlattice, with a monotonous decrease from 0.28 to 0.19 m m / s from y = 26,~ to y = 4A, while the sign of V~, is negative, as consistently deduced from angular variations and from experiments in an external field (fig. 2). These values are consistent with experimental data and with theoretical calculations, which show a correlation of the magnitude of ~ , with the departure of c/a from the ideal value (1.633) [25-27].

At low temperatures (4.2 K), all the M6ssbauer spectra can be decomposed into a non-magnetic subspectrum (identical to the room-temperature spectrum) and a magnetically split contribution (fig. 2). It is systematically observed that the relative proportions of the two subspectra only de-

22 M. Maurer et al. / Hexagonal Fe / Ru superlattices with short periodicity

100.0

99.0

a 0 0 . 0

Z (D 99.5 H 03 (.0

99 .0

(.D Z o~ 100.0

99.8

99 ,6 _

- 4

Fe (Z)Ru(26) ~ TETA=O

T=L. 2K

O0 -2. O0 O. O0 2. O0 4.

VELOCITY( MM/S )

Z C3

co U)

2E U3 Z

100.0 ~ l ~ ~ l ~ ~ ~ , ~ l ~ h l ~ ~1~ ~

99.8

99.6

tO0.0 ~ ~ ~ j ~ ~ ~

99.8

99.6

tO0.O

99,6

- q I O 0 l 3 , 5 0 O. O0 3ll 50 7 , O0

VELOCITY( MM/S )

Fig. 2. ~TFe M6ssbauer spectra of (0001) superlattices: (a) Fe(4 ,~) /Ru(26 A), with the angle 0 = 0 between the superlattice c axis and the y-ray propagation axis; (b) same as (a) except 0 - 54°: (c) same as (a), in a perpendicular external magnetic field of 48kOe at 4.2K. Notice the lack of magnetization at the Fe sites; (d) temperature dependence in the Fe(21 ,~)/Ru(42,~). Notice the moderate decrease in the magnetic splitting up to 300K and the small fraction of the non-magnetic Fe sites.

pend on the Fe thickness x. Actually, when x < o

8 A, the magnetic fraction is practically vanishing. The lack of any magnetic moment and even of any significant susceptibility is nicely demon- strated by a M6ssbauer experiment in an external magnetic field of 48kOe at 4.2 K. In this experiment, the hyperfine field which is measured at Fe nuclei is exactly the external field (48kOe), thus excluding any spontaneous magnetism (even with a small moment) and also any magnetic fluctuations on a typical time scale of 10 7 s. This observation is at variance with the behaviour of the F e / A g [28, 29] or F e / W interfaces [30]. Beyond this limit of about 4Fe monolayers, the magnetic fraction increases like 1 monolayer per monolayer. It is remarkable that the magnetic subspectrum is splitted by an hyperfine field of 337kOe, i.e. very close to the value in pure bcc Fe (343kOe), thus demonstrating that Fe atoms are in a so-called high-spin state, with a magnetic moment reaching about 2 .1#B/Fe atom. This

value is consistent with the theoretical predictions [7, 31, 32]. The relative intensities of the lines from the magnetic sextet indicate that the moments are always aligned in the plane of the film. At this stage, we must mention the likehood of the magnetic behaviour of Fe in hcp (0001) F e / R u superlattices, with the results obtained independently by Liu and Bader [33] for a single F e / R u (0001) interface; in both experiments, it is found that the magnetic interface is sharp, albeit at 2 planes apart from the chemical interface. The magnetism of the thicker superlattice Fe (21A) /Ru(42A) is perfectly in line with data from thinner layers. Obviously, the magnetic subspectrum overwhelms the non-magnetic one. The temperature dependence of the hyperfine field points out to a Curie temperature well above 300K. A preliminary analysis of the T-dependence does not reveal any departure of the spin waves excitations from a T 3/2 behaviour, except for a softening of a factor 3 of the stiffness


constant as compared to bulk bcc-Fe. Some disorder results in a broadening of the hyperfine field distribution upon increasing the temperature.

6. Bulk magnetic properties

Consistently with M6ssbauer spectroscopy, the magnetization measurements reveal the absence of spontaneous magnetization for Fe layers thinner than 4 monolayers, and a subsequent increase of around 2/z B per atom in the additional layer. This evaluation is slightly less accurate than from the hyperfine fields, due to paramagnetic impuri- ties in the substrate. There are also superpara- magnetic contributions with total moments of 20 to 40/z B whose origin remains unclear. For a constant Fe thickness (x = 12A), the shape of hysteresis cycles strongly depends on the Ru thickness [33] with a non-monotonous evolution from a soft (y = 26 A) to a hard behaviour (y = 6 A) [34]. The magnetic anisotropy of some superlattices has been studied. The expected uniaxial anisotropy is not strong enough to counteract the dipolar energy due to the thin film structure, resulting in an in-plane alignment of the Fe moments. A more interesting effect concerns the in-plane anisotropy, which has been studied for one sample Fe(12 ,~) /Ru(6A) by rotating the sample in an external field. Let the anisotropy energy be written as: aE(O)/O0 = )~Aicos(2iO) , with i = 2, 4 and 6. At e.g. 4.2 K, the coefficients are respectively A : = 1.8erg, A 4 =0 .45erg and A~,= 0.2erg. At 300K, the term of order 2 de- creases by 50% whereas the other terms do not change. These findings are another proof for the 6-fold symmetry in the Fe layers.

As a matter of conclusion, the magnetic measurements of F e / R u superlattices permit to es- tablish that the 2 Fe layers at the interface with Ru are absolutely non-magnetic whereas the next ones are fully polarized, with a moment of 2 . 1 t z J F e . This result must be related to a volume expansion exceeding 10% for the Fe layers, due to the pseudomorphous growth with Ru. The

absence of Fe magnetism at the interfaces, which is significant owing to the sharpness of the interface at a monolayer scale, is certainly due to a strong hybridization of the Fe bands with the broader bands of Ru. Angular photoemission experiments are being carried out to clarify this point.

7. Fe growth on (1122) Ru and superlattices of the same orientation

We have recently observed a new epitaxy of Ru (1122) on (1702) sapphire, with flat and single crystalline surfaces [12]. Owing to the pseudomorphous growth of Fe on this Ru surface, we have been able to obtain (117,2)Ru/(ll?,2)Fe superlattices, with an interface quality as good as with the other orientation. This is a remarkable feature, reminding that the (117,2) plane is not dense in the hcp structure. A preliminary evaluation of the interplanar distance versus the Fe and Ru thickness, permits to reach the specific volume per Fe atom in this new phase, i.e. 12.8A 3 [12]. This value is suprisingly comparable with the volume in (0001) superlattices, as a possible sig- nature of a metastable expanded Fe phase, likely related to a high-spin state. The symmetry lowering which is implied in such a growth, as well as the different chemical environments at the F e / R u interfaces are the aim of current investigations.

8. Conclusions

From a systematic study of the crystalline structure and of the local environment around Fe, we have been able to conclude that the (0001) F e / R u superlattices have an hexagonal structure, in which the specific volume of Fe is expanded by more than 10% as compared to equilibrium phases. The F e / R u interfaces are very sharp; correlatively, the magnetic interfaces is also sharp but located at two planes apart inside the Fe layer. The magnetic Fe sites carry a large mo-

24 M. Maurer et al. / lfexagonal Fe / Ru superlattiees with short periodicity

ment of 2.1#t ~. These results are in excellent agreement with theoretical predictions [7, 31, 32] and the fact that the expanded Fe phase is observed for two different pseudomorphous epitax- ies, i.e. (0001) and (1122), strongly supports Kfibler's prediction [7] that an hcp Fe phase could correspond to a metastablc state.

References

[I] D. Bagayoko and J. Callaway, Phys. Rev. B 28 (1983) 5419.

[2] ('.L. Fu, A.J. Freeman and T. Oguchi, Phys. Rev. Lctt. 54 (1985) 2700.

[3] C.L. Fu and A.J. Freeman, Phys. Rev. B 35 (1987) 925. [4] R.F. Willis, J.A.C. Bland and W. Schwarzacher, J. Appl.

Phys, 63 (1988) 4051. [5] V i . Moruzzi, P.M. Marcus and J. Kiibler, Phys. Rev. B

39 (1989) 6957. [6] D.I.C. Pearson and J.M. Williams, J. Phys. F 9 (1979)

1797,

[7] .1. Ki.ibler, Solid State Commun. 72 (1989) 631. [81 M. Maurer, J.C. Ousset, M.F. Ravet and M. Piecuch,

Europhys. l,ett. 9 (1989) 803. [9] [).L. Williamson, S. Bukshpan and R. lngalls, Phys. Rev.

B 6 (1972) 4194. [Ill] L.J. Sehwartzendruber and B. Sundman, Bull. Alloys

Phase Diag. 4 (1983) 155. [111 J.P. Sanchez, M.F. Ravet, M, Piecuch and M. Maurer,

ttyperfine Inleraetions (Proc. ICAME '89, B udapes t ) t o appear.

[12] M. Piccuch, M.F. Raver and M. Maurer, MRS Proc. Spring Meeting 1990, San Francisco (to appear).

[13] A. Dc Andres, M. De Santis, D. Raoux, M. Maurer, M.F, Ravct and M. Piecuch, Proc. X-ray Studies by Syn- chrotron Radiation 11, Roma, 1990. D. Raoux ct al., to appear.

[14] G. Tourillom E. Dartyge, A. Fontaine, M. Lcmonnier and F. Bartol, Phys. Lett. A 121 (1987) 251.

[15] W.G. Mofl'at, Handbook of Binary Phase Diagrams (Gcnium, Schenectady, 1988).

[16] A.R. Miedema and J.W.F. Dorleijn. Surface Sci, 95 (1979) 447.

[17] J.C. Hamilton, Phys. Roy. B 42 (1979)989. [18] F. Gautier and D. Stoeffter, J. Magn, Magn. Mat. 93

(1991) 10. [19] S. Ossicini, R. Memeo and F. Ciccacei, J. Vac. Sci.

Technol, A3 (1985) 387. [20] M, Farlc and K. Baberschke, Phys. Rev. Lett. 58 (1987)

511. [21] W. Diirr. T. Woike, T. Beicr and D. Pescia, J. de Phys. 49

(1988) C8-1615. [22] M. De Santis ct al., in preparation. [23] F. Baud¢lct et al., in preparation. [24] A. Bourret and J.L. Rouvi~re, MRS Proc. Spring Mcct

ing 1990, San Francisco (to appear). [25] G. Wortmann and D.L. Williamson, Hyperline Interac-

tions I (1975) 13. [26] M. Piecuch and C. Janot, Hypcrfine Interactions 11 (1981)

13. [27] P. Blaha, K. Schwarz and P.[1. Dederichs, Phys. Rev. B

37 (1988) 2792. [28] Z.O. Oiu, S.[I. Mayer, CJ . Guttierrez, II. Tang and J.('.

Walker, Phys. Rcv. Lctt, fi3 (1989) 1649. [29] N .C Koon, B.T. Jonkcr, F.A. Volkening, J.J. Krchs and

G.A. Prinz, Phys. Rev. Lett. 59 (1987) 2463. [30] M. Przybylski and U. Gradmanm Phys. Rev. Lctt. 59

(1987) 1152. [31J D. Knab and ('. K6nig, ,|. Magn. Magn. Mat. 93 (It)91}

398. [32] M.O. Selm¢ and P. Pecheur, J. Magn, Magn. Mat. 93

( 1991 ) 285. [33] ('. Liu and S. Bader, Phys. Rev. B 41 (199(I} 553. [34] M. Maurer, J.C. Ousset, M. Piecuch, M.F. Ravel and J.P.

Sanchez, Proc. MRS Spring Meeting 1989, San l)iego, MRS Sylnp. Proc. 151 (1989)99.

Magnetism and structure in hexagonal Fe/Ru superlattices with short periodicity

Documents