Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://danp.sinp.msu.ru/LIIWM/PhysRevB58_1998_8960Chu.pdf Äàòà èçìåíåíèÿ: Wed Oct 24 00:16:16 2007 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 22:19:37 2012 Êîäèðîâêà:

PHYSICAL REVIEW B

VOLUME 58, NUMBER 14

1 OCTOBER 1998-II

Structure of Ni overlayers on bcc Fe,,100...
A. V. Mijiritskii,* P. J. M. Smulders, V. Ya. Chumanov, O. C. Rogojanu, M. A. James, and D. O. Boerma
Materials Science Centre, University of Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands Received 16 March 1998 The structure of Ni layers epitaxially grown on bcc Fe 100 was studied by various techniques. After the growth of up to 3 monolayers ML , Ni was found to be in a bcc phase with the bcc Ni 100 bcc Fe 100 epitaxial relationship and with the in-plane lattice parameter close to that of the underlying Fe. When the Ni layer thickness exceeded 3 ML, a structural transition was observed. In the literature it has been reported that the intermediate Ni phase formed upon the transition has an unknown structure which gradually changes to a final fcc Ni 110 structure after the growth of about 200 ML and subsequent annealing. Here we find that this intermediate phase is in fact also fcc Ni 110 with a structure consisting of small domains. It is concluded that the small domain size gives rise to reflection high-energy electron diffraction patterns which can be misinterpreted as being due to a different structure. The intermediate as well as the final Ni structure contains four orientations of fcc Ni domains with the Ni 110 planes parallel to the bcc Fe 100 surface and the Ni 211 crystallographic directions along Fe 110 axes. The origin and consequences of this epitaxial relationship are discussed. S0163-1829 98 08838-9 I. INTRODUCTION

The Ni/bcc Fe 100 superlattice has been the subject of theoretical and experimental studies during the last decade. In spite of this, a number of features, both of the crystallographic structure and of the magnetic properties of this system, remain unclear. A bulk Ni crystal has the fcc structure and no other phases have been observed at normal temperature and pressure. Nevertheless, it is possible to stabilize metastable phases by means of epitaxial growth of thin films on appropriate templates.1­4 It has been found that the first few monolayers of Ni grown on a bcc Fe 100 template exhibit a metastable bcc phase. According to theoretical calculations, metastable bcc Ni is nonmagnetic and has a lattice parameter equal to 2.773 å.5­7 However, in thin epitaxial Ni films on bcc Fe 100 the in-plane lattice spacing is expanded to approximately that of Fe 2.866 å to recover the mismatch between the bcc Ni and bcc Fe unit cells.8­10 The expansion causes a contraction of the interplanar spacing in the Ni overlayer. Such a contraction has been observed recently by Kamada and co-workers11,12 who determined the interplanar spacing in 3-ML-thick Ni layers to be 1.35 å. Upon reaching a critical thickness, the Ni overlayer undergoes a reconstruction, forming an intermediate Ni phase. On the basis of the theory of homogeneous strains,13 the critical thickness of the bcc Ni overlayer was calculated to be 25 ML.9 In fact, the critical thickness was observed14,15 to be only 3 ­ 8 ML. As stated in Ref. 14, this discrepancy may be caused by the presence of surface defects. The structural changes cause additional streaks in ( 1 1 ) reflection highenergy electron diffraction RHEED 9,11 and low-energy electron diffraction LEED 10 patterns. The patterns indicate a c (2 2 ) -like surface reconstruction.9,10 However, the observed patterns differ slightly from normal c (2 2 ) patterns and, for RHEED, they depend on the angle of incidence of the electron beam. After a short annealing of a sample with an 20-ML-thick Ni layer, Wang, Jona, and Marcus10 observed a change in the LEED pattern. The pattern was interpreted as a superposition of a bcc Ni 100 mash and two fcc
0163-1829/98/58 14 /8960 7 /$15.00 PRB 58

Ni 100 mashes rotated about 10° on either side of the bcc mash. Recently, the structure and magnetic properties of metastable Ni/bcc Fe 100 multilayers have been studied by Kamada and co-workers11,12 using RHEED, x-ray diffraction XRD , and superconducting quantum-interference detector magnetometry SQUID . On the basis of their observations, the authors proposed three structural phases of a Ni layer deposited onto the bcc Fe 100 template: i a 3-ML-thick Ni layer has a distorted bcc phase as described above; ii when the Ni layer thickness exceeds 3 ML, a reconstruction takes place leading to an intermediate c (2 2 ) -like structure, as revealed by RHEED; iii upon further growth, the intermediate Ni changes its structure gradually from c (2 2 ) -like to fcc consisting of twinned Ni 110 domains rotated by 9.74° with respect to Fe 100 directions, giving rise to a four-domain structure. The authors proposed that 211 crystallographic directions of fcc Ni 110 domains coincide with 110 directions of the bcc Fe 100 template. The formation of fcc Ni domains was also observed after a short annealing of an intermediate Ni layer with a thickness of 200 å at about 250 ° C. The fcc Ni 110 211 bcc Fe 100 110 epitaxial relationship found by Kamada and co-workers from interpretation of RHEED patterns has not been confirmed yet by a more direct measurement. Moreover, the structure of the intermediate Ni layer as well as the nature of the epitaxial relationship mentioned above is still not understood. Besides, the peculiarities of the magnetic properties11,14 ­ 18 observed in metastable Ni/ bcc Fe layers still require clarification. In this paper we present an extensive analysis of Ni/bcc Fe 100 bilayers and a possible model to explain the transformation occurring in the Ni overlayer.
II. EXPERIMENTAL DETAILS

The ultrahigh vacuum UHV molecular beam epitaxy MBE setup consists of three interconnected chambers, all with a base pressure of better than 5 10 10 mbar. MgO single crystals were cleaved in air and subsequently annealed
8960 © 1998 The American Physical Society


PRB 58

STRUCTURE OF Ni OVERLAYERS ON bcc Fe 100

8961

FIG. 1. Geometry of the RBS/Channeling experiments.

in the preparation chamber in oxygen P O2 1 10 6 mbar for 12 hours at a temperature of about 670 ° C to remove adsorbates and surface imperfections. The sample surface structure was analyzed in the main chamber by RHEED with a 15-keV electron beam. RHEED patterns were digitized by a charge-coupled device CCD camera so that the distances between RHEED streaks could be precisely measured. The in situ x-ray photoelectron spectroscopy XPS studies were performed in the analytical chamber equipped with a CLAM2 hemispherical analyzer. In the experiments, Mg K radiation was used to avoid overlapping of the Ni LMM Auger series with the Fe 2 p doublet. From RHEED and XPS measurements it was found that, after annealing, the MgO substrates had well-ordered 100 surfaces with no contamination. The sample growth was carried out in the main chamber. The 200 ­ 300-å-thick Fe layers were deposited at a substrate temperature of 250 ° C. Because the surface of the asdeposited layers was found to be rough, the samples were subsequently annealed at 450 ° C in the preparation chamber under UHV conditions about 8 10 10 mbar for 10 min.19 An ordered surface was obtained, as observed by RHEED. The shape and the binding energy of the Fe 2 p doublet in XPS showed that Fe was in the metallic state both before and after annealing. Afterwards, the Fe layers were covered with Ni in UHV at either ambient temperature or at 150 ° C. No interdiffusion or oxide formation was detected by XPS. The thickness of the Ni layers was varied over a range of 50 ­ 200 å. To take RHEED angular scans for some of the samples, the Ni deposition was interrupted after deposition of a few ML. Rutherford backscattering spectroscopy RBS/ Channeling , atomic force microscopy AFM , scanning electron microscopy SEM , and XRD measurements were carried out ex situ. In RBS/Channeling, a 1-MeV He beam was used. The backscattered particles were detected at an angle of 165° . The sample to be investigated was mounted on a two-axis goniometer with a long-range reproducibility and accuracy of better than 0.1° . The orientation of the samples was established by performing RBS/Channeling on a bare part of the MgO substrate. During the measurements, the pressure did not exceed 5 10 8 mbar. A detailed analysis of the measured RBS spectra was done using the programs RUMP Ref. 20 and RBSIM.21 The thickness of Ni oxide formed at the surface while transferring the sample through air to the RBS/Channeling setup was determined from RBS to be about 10 å. In the RBS spectra, the Fe and Ni peaks were overlapping so that a deconvolution procedure

FIG. 2. Azimuthal RBS scans. The azimuthal angles of 0 ° and 90° correspond to MgO 010 and MgO ¯ 00 crystallographic 1 directions, respectively. The polar angle is fixed at 76° . The scans are normalized to a random spectrum yield. The positions of dips expected for the crystallographic structures described in the text are indicated.

had to be used to extract correct backscattering yields. The XRD measurements were carried out in a Philips X'PertMRD system in thin film configuration. X-ray reflectivity measurements were performed to determine the sample structure, using Cu K radiation. A Ni filter was used to remove the Cu K radiation. The XRD spectra were analyzed using a fitting procedure described elsewhere.22 The AFM analysis was performed using a Nanoscope II Digital Instruments setup. The SEM studies were carried out with a Philips XL30S microscope.
III. RESULTS

In Fig. 1, the geometry used in the RBS/Channeling measurements is indicated. The polar angle is the angle between the sample surface and the incident beam. The aziis measured from the MgO 010 muthal angle crystallographic direction in the surface plane. Figure 2 shows azimuthal scans of the RBS yield of Mg, Fe, and Ni measured for the Ni 200 å /Fe 200 å /MgO sample. The azimuthal scans were acquired at a fixed polar angle, 76° . Figure 3 displays polar scans of the RBS yield for the same sample recorded at an azimuthal angle of


8962

A. V. MIJIRITSKII et al.

PRB 58

FIG. 4. Four possible fcc Ni 110 domains black dots on the bcc Fe 100 mesh open circles .

FIG. 3. Polar RBS scans. to the surface normal MgO positions of dips expected for For Ni the dip positions are domain orientations given in

The polar angle of 90° corresponds 001 crystallographic direction . The the described structures are indicated. labeled I ­ IV, in accordance with the Fig. 4.

0° MgO 001 direction . Upon measuring, the sample is rotated and the trajectory of the incident beam crosses crystallographic planes of the sample studied. When the incident beam propagates along crystallographic planes of the crystal, channeling dips in the RBS scans are observed. The positions of these dips correspond exactly to real-space axial or planar directions of the crystalline layer. As expected, the bcc Fe lattice was found to be rotated by 45° around an axis perpendicular to the 001 surface plane with respect to the MgO lattice. This epitaxial relationship leads to a small 3.7% mismatch between interatomic distances along MgO 100 and Fe 110 crystallographic directions. In our previous studies we found that the lattice strain caused by this mismatch in a 200-å-thick Fe layer is relaxed.23 The positions of all the channeling dips found in Ni in both azimuthal and polar scans can be explained assuming that Ni is in the fcc phase and that the epitaxial relationship between fcc Ni and bcc Fe is Ni 110 R 9.74° Fe 100 . In this case, four domains for fcc Ni can be expected, where two pairs of mutually perpendicular domains are rotated by 9.74° with respect to the Fe 100 crystallographic direction see Fig. 4 . In Figs. 2 and 3, the positions of planar and axial channeling dips expected for this structure are indicated. As can be seen, all dips observed are at such predicted

positions. The determined orientation of the Ni lattice and the presence of domains were confirmed by recording a polar scan along a Ni 110 plane at a fixed azimuthal angle of 35.26° not shown . In this scan, channeling dips were also observed at the positions corresponding to the Ni structure described above. To verify the results obtained, we performed XRD measurements on a Ni 200 å /Fe 200 å /MgO sample. A distinct peak was observed at 2 Bragg 76.4° . Within the experimental uncertainty, this corresponds to a Bragg reflection at 76.37° for an interplanar distance of 2.49 å, what is the distance between 110 planes in an fcc Ni crystal. These data indicate unambiguously that the Ni 110 plane is parallel to the sample surface and, thus, to the Fe 100 plane. RHEED patterns for the electron beam along the Fe 110 direction are presented in Fig. 5. An as-prepared bcc Fe layer reveals a well-ordered ( 1 1 ) surface Fig. 5 a . A 4-åthick Ni overlayer 3 ML, assuming a 1.35 å interlayer spacing shows the same surface structure as the underlying Fe Fig. 5 b . This indicates that Ni forms an epitaxial bcc layer with in-plane lattice dimensions which are close to those of the Fe template. No structural differences between the Ni overlayers grown at about room temperature and at 150 ° C were found. At a Ni thickness of 8 å, a significant broadening of the RHEED streaks and the appearance of additional features is observed see Fig. 5 c . These changes point to a transformation occurring in the Ni layer upon increasing thickness. The additional features make the RHEED pattern similar to that of a c (2 2 ) surface reconstruction. However, the observed RHEED pattern differs from a c (2 2 ) pattern, since the separation between the additional streaks is shorter by 6.0 0.8% than the half-order streak separation expected for a normal c (2 2 ) pattern. To study the origin of the streaks, we recorded angular RHEED scans. The sample was rotated around an axis perpendicular to the surface plane to vary the azimuthal angle of the incident electron beam, and RHEED patterns were taken at 3 ° intervals. Figure 6 shows a projection of the reciprocal lattice of four fcc Ni 110 domains onto the bcc Fe 100 plane. A schematic representation of the RHEED pattern of Fig. 5 c is given in Fig. 7 a . The field of the CCD camera view is indicated with a dashed line. Figure


PRB 58

STRUCTURE OF Ni OVERLAYERS ON bcc Fe 100

8963

FIG. 6. Projection of the reciprocal lattice of four fcc Ni 110 domains as indicated in Fig. 4 onto the bcc Fe 001 plane. The four domains are indicated by different symbols. The black symbols correspond to the spots due to the diffraction from the atoms in the topmost Ni layer. The white symbols correspond to the spots due to the diffraction from Ni atoms in the second layer. FIG. 5. RHEED patterns taken with the electron beam along ¯ Fe 11 0 : pattern a corresponds to the as-prepared bcc Fe 001 surface; pattern b corresponds to a 4-å-thick Ni overlayer on the bcc Fe 001 surface; pattern c corresponds to an 8-å-thick Ni overlayer on the bcc Fe 001 surface.

7 b displays the observed angular dependence of the firstorder streak separation D1 and the separation D2 of the additional streaks. The diffraction spot ``A'' see Fig. 6 is in fact an overlap of spots from two fcc Ni domains with a mutual angle of 19.48° as shown in Fig. 4. Along the Fe 110 direction, spot ``A'' causes first-order diffraction streaks with separation D1 corresponding to the fcc Ni structure. In this direction, bcc Ni has approximately the same within 0.3% first-order streak separation. When turning away from the Fe 110 direction over an angle , the distance D1 between the first-order streaks decreases, reaches a minimum at about 10° , and increases again. The initial decrease is the result of the projection of spot ``A'' onto the current direction at which the pattern is acquired. Since the ``O'' ­ ``A'' and ``O'' ­ ``B'' distances in the reciprocal lattice projection are equal see Fig. 6 , the increase in the first-order streak separation beyond 10° is due to the proximity of diffraction spot ``B.'' Note, that for 9.74° the electron beam is incident in a direction halfway between the spots ``A'' and ``B.'' In contrast to the angular dependence of D1, the separation D2 between the additional streaks increases gradually. Keeping in mind that we deal with transmission RHEED patterns, the weak additional

streaks ``X'' could represent a cut of the extended diffraction spot ``C'' see Fig. 6 , which is a first-order diffraction spot along the fcc Ni 100 directions. The projection of diffraction spot ``C'' onto the Fe 110 direction results in streaks with a separation which is shorter by 5.7% than the halforder separation expected for a correct c (2 2 ) reconstruction. In our measurements the separation D2 is shorter by 6.0 0.8% than that expected for c (2 2). AFM and SEM studies of both as-deposited and annealed ( 200 ° C) Ni 200 å /Fe 200 å /MgO samples revealed domains covering the sample surface. The size of the domains was in the range of 600 ­ 1700 å. Detailed analysis of the AFM and SEM images showed that some of the domain boundaries intersect at typical angles of about 20° , 35° , 55° , or 90° . Considering four fcc Ni domains with the 110 surface orientation, the observed domain boundary directions can be interpreted as low-index crystallographic planes. The mutual orientation of the domains and their orientation with respect to macrosteps on the MgO substrate running along MgO 100 and MgO 110 are in good agreement with the RBS/Channeling data discussed above. Sometimes, holes in the Ni layer are formed at junctions of the domain boundaries. The size of the holes was observed to be in the range 200 ­ 400 å with AFM. An XPS analysis of the as-deposited samples indicated a small amount of near-surface Fe corresponding to about 5 2 % of the surface area. The Fe signal in XPS could originate from the underlying Fe layer exposed at the Ni domain boundaries.


8964

A. V. MIJIRITSKII et al.

PRB 58

FIG. 8. A bccfcc martensitic transition in Ni: a a bcc Ni unit cell with the lattice parameter equal to that of bcc Fe 2.866 å ; b the Bain transformation; c the atomic positions of Ni atoms black dots after the bccfcc transition of a Ni layer on bcc Fe 100 open circles . The grey dots indicate Ni atoms, which positions stay unaffected upon the transition.

FIG. 7. a schematic drawing of a measured RHEED pattern corresponding to a 8-å-thick Ni overlayer on the bcc Fe 001 sur¯ face taken with the electron beam along Fe 11 0 . The meaning of the spots ``A,'' ``E,'' ``O,'' and ``X'' is shown in Fig. 6 and explained in the text. The dashed square depicts the CCD camera field of view; b changes of the first-order diffraction streak separation D1 and the separation D2 of the additional streaks as a function of the azimuthal angle of the RHEED electron beam. IV. DISCUSSION

To explain the results obtained, we propose that the transformation observed in the Ni overlayers on a Fe 100 surface represents a bcc Nifcc Ni martensitic transition. In accordance with that mentioned in the introduction, we assume that before the transition, bcc Ni has approximately the same lattice parameter as bcc Fe.10,14 Upon the transition, the atomic movements can be decomposed into: i a distortion of the bcc Ni in accordance with the Bain transformation24 as shown in Fig. 8 b ; ii a rotation in the bcc Ni 001 plane by an angle of 9.74° as shown in Fig. 8 c . The combination of these two components leads to shifts of the Ni atoms along bcc Ni 110 directions, so that the bcc Ni 001 plane shaded in Fig. 8 a translates into the fcc Ni 110 plane shaded in Fig. 8 b and every fifth bcc Ni 230 plane stays intact upon its translation into a fcc Ni(¯ 23) plane. Figure 9 indicates the 2 coincidence sites crosses of the bcc Fe open circles and fcc Ni black dots lattices upon the transition. The driving force for the transition is the free energy released by the change of the crystal structure in order to reduce its elastic distortion and to adopt a more stable crystal structure. This

energy increases linearly with the thickness of the Ni layer. At the same time, the transition costs energy because atomic bonds between Ni and Fe atoms are broken. The transition occurs when the elastic energy exceeds the energy required for breaking the atomic bonds and when the relevant kinetic barriers can be overcome. This can happen when the Ni layer reaches a critical thickness. The transition will proceed in such a way that the bond breaking is reduced to a minimum. In other words, the epitaxial relationship between the fcc Ni overlayer structure and the underlying bcc Fe crystal must lead to a maximum coherence. In the system being discussed, the epitaxial relationship is fcc Ni 110 211 bcc Fe 100 110 . This relationship leads to a semicoherent Fe-Ni interface, as depicted in Fig. 9. We propose that the nucleation starts at the sites where two

FIG. 9. The coincidence sites crosses of the bcc Fe open circles and fcc Ni black dots lattices. Every fifteenth Ni-Fe pair at the Ni/Fe interface stays unaffected upon the transition.


PRB 58

STRUCTURE OF Ni OVERLAYERS ON bcc Fe 100

8965

Ni atoms stay in place, as in Fig. 8 c . Consecutively, the adjacent cells undergo the transition. Nucleation may occur at many sites. Initially, a structure with small domains will be formed. At higher temperatures larger domains will grow in size at the cost of smaller domains. The arial density of atoms in fcc Ni 110 planes is smaller than in strained bcc Ni 100 planes by about 7%. For a fcc Ni 110 layer consisting of small domains this density is even more reduced. This implies that during the transition Ni atoms are expelled from the fcc layer formed. The holes observed with AFM in annealed transformed Ni layers may be caused by a reduction of the total length of domain boundaries as a result of the formation of larger domains upon annealing. The proposed model is in good agreement with experimental results reported in the literature.9,12,25 ­ 27 The decrease in coverage upon growth of a Ni layer on Fe 100 was also found indirectly by Heinrich and co-workers as a decrease in the period of RHEED intensity oscillations.14,25 Our conclusion that the intermediate Ni consists of fcc Ni domains is supported by glancing incidence extended x-ray absorption fine structure EXAFS measurements.27 The first- and second-shell Ni-Ni distances in a 37-ML-thick Ni layer were found to be compatible with the fcc structure. The amplitude of the third and fourth shells indicated disorder of the lattice which could be due to the presence of small domains. Kamada and Matsui reported that annealing of the sample leads to a decrease of intensity of the additional RHEED streaks and to a development of a pure fcc Ni 110 pattern.12 According to their RHEED patterns, it seems the annealing was also accompanied by a decrease of the extensions of all the streaks present. The reduction of the streaks extensions could be a consequence of growth of the domain sizes. Domains with sizes in the order of the coherence length of the electrons in the RHEED beam will cause a pattern with broadened streaks. If the domain size exceeds the coherence length, a pattern with sharp streaks will develop. The weakening of the additional streaks upon annealing can be due to a sharpening of the RHEED streaks ``C,'' which have their maximum intensity along Ni 100 directions see Fig. 6 . In the same fashion, the reported12 changes in the separation between the pair of streaks corresponding to spot ``E'' of the reciprocal lattice could be caused by the presence of extended spots ``F'' and ``G'' see Fig. 6 . Along the Fe 100 direction, the fcc Ni first-order RHEED diffraction spots ``C'' and ``D'' from two mutually perpendicular domains give projections which are very close to the bcc Ni or bcc Fe first-order spot see Fig. 6 . Upon structural transition, these spots may overlap and can create a broad single streak near the previously observed bcc Ni streak position. This may be a misleading factor in a correct interpretation of the development of the RHEED patterns due to the transition. The structural peculiarities of a Ni overlayer on bcc Fe 100 are in partial agreement with the magnetic properties reported in the literature.9,14,18 Heinrich and co-workers9,14 found that a transformed Ni overlayer on a thin bcc Fe 100 layer on a vicinal Ag 100 substrate exhibits an in-plane anisotropy, which is much larger than the anisotropy of a thin layer of Fe, and this anisotropy possesses a fourfold symme-

FIG. 10. Illustration of how 111 vectors in four different fcc Ni 110 domains add to two mutually perpendicular vectors A and B. The crystallographic aspects of the structure are explained in Fig. 4.

try. It was also observed that upon transformation of the Ni overlayer, the easy axes rotate from 100 to the 110 directions of bcc Ni 100 . The orientation of the easy axes in transformed Ni coincide with 111 crystallographic directions of fcc Ni domains see Fig. 4 and with the directions of a fraction of the domain boundaries observed in our AFM and SEM measurements. The fourfold symmetry of the inplane anisotropy observed cannot be recognized easily in the four-domain structure of the fcc Ni layer. However, the vector sums of the 111 directions in different domains as indicated in Fig. 10 are in mutually perpendicular directions. The significant enhancement of the anisotropy itself may be due to the presence of domain boundaries.
V. CONCLUSION

We propose that the transformation observed in Ni overlayers on bcc Fe 100 represents a bccfcc martensitic transition in Ni, meaning that fcc Ni is formed directly from the bcc phase and no other intermediate phase exists. The intermediate Ni layer is a mosaic of four fcc Ni domains. The c (2 2 ) -like reconstruction observed in RHEED patterns is probably an artifact due to the existence of small Ni domains. Upon the martensitic transition, the fcc Ni 110 planes arrange parallel to the bcc Fe 100 surface. This relationship can be understood from the geometry of the bcc and fcc unit cells and is a direct result of the martensitic transition, i.e. it is not the result of an epitaxial growth accompanied by gradual shifts of Ni atoms throughout the whole thickness of the layer. To match to the bcc Fe 100 mesh, the fcc Ni 110 domains are rotated in the surface plane by 9.74° with respect to the bcc Fe 100 crystallographic directions so that fcc Ni 211 bcc Fe 110 . The size of the domains of a 200-å-thick Ni layer was observed to be in the range 600 ­ 1700 å.
ACKNOWLEDGMENTS

We would like to thank Professor T. Hibma, Professor L. Niesen, and F. C. Voogt for useful discussions. We are also grateful to Dr. P. M. Bronsveld and J. W. Hooijmans for the SEM data acquisition. One of the authors V.Ya.Ch. gratefully acknowledges the hospitality met at the University of Groningen. His stay at the University of Groningen was made possible by financial support from NWO Dutch Organization for Scientific Research .


8966

A. V. MIJIRITSKII et al.

PRB 58

*Author to whom correspondence should be addressed. FAX: 31

50 363 3471. Electronic address: A.Mijiritskii@phys.rug.nl Permanent address: Institute of Nuclear Physics, Moscow State University, Moscow 119899, Russia. 1 M. Matsui, S. Mitanai, and A. Kida, in New Functionality Materials, edited by T. Tsuruta Elsevier Scientific Publishers, Amsterdam, 1993 , Vol. C, p. 125. 2 A. Kida and M. Matsui, J. Phys. Soc. Jpn. 65, 1409 1996 . 3 D. Li, M. Freitag, J. Pearson, Z. Q. Qiu, and S. D. Bader, Phys. Rev. Lett. 72, 3112 1994 . 4 B. Heinrich, Z. Celinski, J. F. Cochran, W. B. Muir, J. Rudd, Q. M. Zhong, A. S. Arrott, K. Myrtle, and J. Kirschner, Phys. Rev. Lett. 64, 673 1990 . 5 V. L. Moruzzi, P. M. Marcus, K. Schwarz, and P. Mohn, Phys. Rev. B 34, 1784 1986 . 6 V. L. Moruzzi, Phys. Rev. Lett. 57, 2211 1986 . 7 V. L. Moruzzi and P. M. Marcus, Phys. Rev. B 38, 1613 1988 . 8 B. Heinrich, A. S. Arrott, J. F. Cochran, C. Liu, and K. Myrtle, J. Vac. Sci. Technol. A 4, 1376 1986 . 9 B. Heinrich, S. T. Purcell, J. R. Dutcher, K. B. Urquhart, J. F. Cochran, and A. S. Arrott, Phys. Rev. B 38, 12 879 1988 . 10 Z. Q. Wang, Y. S. Li, F. Jona, and P. M. Marcus, Solid State Commun. 61, 623 1987 . 11 Y. Kamada, M. Matsui, and T. Asada, J. Phys. Soc. Jpn. 66, 466 1997 . 12 Y. Kamada and M. Matsui, J. Phys. Soc. Jpn. 66, 658 1997 . 13 A. A. Chernov, in Modern Crystallography III-Crystal Growth, edited by B. K. Vainshtein Springer-Verlag, Berlin, 1984 , p. 99. 14 B. Heinrich, J. F. Cochran, A. S. Arrott, S. T. Purcell, K. B.

15

16

17

18

19

20

21

22

23

24

25

26

27

Urquhart, J. R. Dutcher, and W. F. Egelhoff, Jr., Appl. Phys. A: Solids Surf. 49, 473 1989 . N. B. Brookes, A. Clarke, and P. D. Johnson, Phys. Rev. B 46, 237 1992 . C. J. Gutierrez, M. D. Wieczorek, Z. Q. Qiu, H. Tang, and J. C. Walker, J. Magn. Magn. Mater. 93, 369 1991 . J. A. C. Bland, R. D. Bateson, A. D. Johnson, B. Heinrich, Z. Celinski, and H. J. Lauter, J. Magn. Magn. Mater. 93, 331 1991 . Q. M. Zhong, A. S. Arrott, B. Heinrich, and Z. Celinski, J. Magn. Magn. Mater. 104-107, 1837 1992 . M. Dynna, J. L. Vassent, A. Marty, and B. Gilles, J. Appl. Phys. 80, 2650 1996 . L. R. Doolittle, Nucl. Instrum. Methods Phys. Res. B 9, 344 1985 . P. J. M. Smulders and D. O. Boerma, Nucl. Instrum. Methods Phys. Res. B 29, 471 1987 . M. A. James, F. C. Voogt, L. Niesen, O. C. Rogojanu, and T. Hibma, Surf. Sci. 402-404, 332 1998 . A. V. Mijiritskii, P. J. M. Smulders, D. O. Boerma, and F. C. Voogt unpublished . A. H. Cottrell, in An Introduction to Metallurgy, edited by A. H. Cottrell Edward Arnold Ltd., London, 1967 , p. 319. S. T. Purcell, A. S. Arrott, and B. Heinrich, J. Vac. Sci. Technol. B 6, 794 1988 . J. Vogel, G. Panaccione, and M. Sacchi, Phys. Rev. B 50, 7157 1994 . D. T. Jiang, N. Alberding, A. J. Seary, B. Heinrich, and E. D. Crozier, Physica B 158, 662 1989 .