Allenspach

JTsgL&
ELSEVIER
-E3
$1
4 ._
Journal of Magnetism and Magnetic Materials 129 (1994) 160-185
Ultrathin films: magnetism on the microscopic scale

R. Allenspach
IBM Research DiGsion, Zurich Research Laboratory, 8803 Riischlikon, Switzerland
(Received 22 July 1993)
Abstract
The basic concepts associated with two-dimensional ferromagnetism are illustrated from an experimentalists viewpoint. Ultrathin epitaxial films are the prototype systems for investigating magnetism in reduced dimensions, and spin-polarized scanning electron microscopy is a powerful technique for such experiments. The topics covered include a discussion on the formation of magnetic domains in these films, the microscopic investigation of spin fluctuations, and the key role that anisotropies play in these systems.
1. Introduction
The art of growing epitaxial metal films has engendered fascinating new topics in materials research. Ultrathin films approaching a single . monolayer represent model systems to investigate physics in two dimensions (2D). One of the fastest growing subjects in this field is ferromagnetism in these reduced dimensions. In the past few years, exciting new effects have been discovered, such as, for example, ferromagnetic-antiferromagnetic oscillations in exchange coupling or giant magnetoresistance. Concurrently with these discoveries, novel techniques have been developed to investigate these ultrathin films with high sensitivity. In particular it has been realized that not only high sensitivity is essential, but also lateral resolution. The reason is that a single-domain state, although its occurrence has been anticipated in thin films for decades, is by no means always the preferred ground state. This lecture tries to give an overview of those physical aspects encountered in ultrathin mag0304-8853/94/$07.00
netic films where the high sensitivity and spatial resolution lead to new insights into 2D magnetism. We employ spin-polarized scanning electron microscopy (spin-SEM, or SEMPA) to investigate the domain patterns in these epitaxial layers with high spatial resolution. We will see that an understanding of domain structures and the ability to determine the magnetization direction within a domain are vital for the interpretation of various properties of a 2D ferromagnet. The paper is organized as follows: Section 2 summarizes the basic theoretical concepts underlying the formation of domains and defines the relevant quantities encountered in subsequent chapters. In Section 3 the technique of spin-SEM is introduced and the experimental setup described. Experimentally observed domain patterns for various epitaxial layers are presented in Section 4. Thermal fluctuations drastically affect the behavior of 2D systems: Their influence on the magnetic response below and above the Curie temperature is explained in Section 5, with emphasis on the conceptual difference between
0 1994 Elsevier Science B.V. All rights reserved
XjDZ0304-8853(93)E0509-B
R. Allenspach /Journal
of Magnetism and Magnetic Materials 129 (1994) 160-185
161
spontaneous and remanent magnetization. Section 6 investigates the special role that magnetic anisotropies play in ultrathin films. The balance between surface and shape anisotropies can lead to a reorientation of the magnetization direction, either with changing film thickness or temperature. This phase transition induces unusual domain patterns which are ascribed to the drastically different 2D behavior of a magnet as compared to 3D.
2. Why do magnetic domains
exist?
The concept of magnetic domains was already postulated at the beginning of the 20th century by Weiss [l]. He was the first to connect two observations that are unique to ferromagnetic materials, namely (i) a huge magnetization as a response to a small external magnetic field, and (ii) a remanent magnetization which may or may not vanish, depending on the magnetic history of the sample. These observations were the key ingredients in the postulate that a ferromagnetic material in zero applied field usually decays into regions with full magnetization, while the direction of this spontaneous magnetization varies from region to region. The effect of a small external field in such a specimen then is to align the magnetization directions of the various domains rather than to induce magnetization on the atomic scale. Some thirty years later, Landau and Lifshitz [2] showed that the subdivision of a 3D-bulk ferromagnetic sample into domains can considerably reduce the magnetostatic energy compared to a single-domain state, which firmly established the theoretical foundation of the earlier postulate. The didactic and introductory purpose of this lecture requires a sketch of the physical ideas behind domain formation in the general 3D case, with the aim of also establishing the essentials for the ultrathin films with their 2D nature. For a more detailed discussion of the basics of domain formation, the reader is referred to one of the many standard textbooks, e.g. Refs. [3] and [4]. The main contributions to the total energy of a
ferromagnet without external field are the exchange energy, the magnetocrystalline anisotropy energy, and the magnetostatic energy, also known as demagnetization energy or shape anisotropy. Ferromagnetism is a phenomenon of quantum mechanical origin. A key role is played by the exchange interaction between electron spins at different atomic sites, which tries to keep adjacent spins aligned parallel in a ferromagnet (and hence produces a spontaneous magnetization). We may therefore attribute an energy term E, = -JS, . Sj to this interaction, where Si,j are the spins of electron i,j and J is the exchange constant (J > 0 for ferromagnets, J < 0 for antiferromagnets). If this succinct description seems to suggest an interaction between the spin magnetic moments, such a conclusion is physically misleading. It merely reflects the tendency of a mathematical formula to obscure what it seeks to illuminate: The exchange interaction is a direct consequence of Paulis exclusion principle, and hence of electrostatic origin. Nevertheless, its consequence is the existence of magnetic long-range order! The magnetocrystalline anisotropy energy describes the coupling of the electron spin to its surroundings, the crystal lattice, by a spin-orbit coupling energy. It results in preferred crystallographic directions along which all the aligned spins (or the magnetization) lie. The general anisotropy energy term can be deduced from the symmetry of the crystal lattice. In its simplest form of one preferred direction, the case of uniaxial anisotropy, it can be described by an energy EA = Ksin20, where K is the anisotropy constant and 0 the angle between the easy anisotropy direction and the magnetization. Because of the quadratic sine-dependence it is immediately clear that at least two orientations are energetically degenerate, i.e. that the sign of the magnetization can be chosen arbitrarily. When we realize that the magnetocrystalline anisotropy is a macroscopic description of the interaction between electron spin and lattice, it intuitively becomes clear that minute changes of the lattice constant will affect the magnetic properties. In particular, if the lattice is strained along a certain direction, there are additional contributions to
162
the anisotropy energy, which are usually summarized as magnetostrictive or magnetoelastic anisotropy energy. For the moment, we do not consider them as a separate entity, since in their simplest form they have the same functional dependence as a magnetocrystalline uniaxial anisotropy does. The third relevant energy term is the magnetostatic energy. Aligning all spins in a bulk sample along the easy magnetization direction determined by the crystalline anisotropy produces a huge magnetic stray field outside the sample. This energy is given by E, = pa/2 x NM2, with M the magnetization vector per unit volume and N a demagnetization constant depending on the exact shape of the sample. In particular, 0GN G 1, and the limiting values are realized in a thin film with magnetization parallel (N = 0) and perpendicular to the plane (N = 11, respectively. This is the reason why this magnetostatic demagnetization energy is also called shape anisotropy. The maximum value of this energy term is taken in the uniform state with all spins parallel. Therefore it is energetically favorable to end up with a spin configuration where all spins are aligned within certain regions, but along different directions from region to region so that the magnetic stray field is cancelled to a great extent. Within these regions - the magnetic domains - the spins point along one of the easy magnetization axes, and only in the narrow transition zones - the domain walls - does the spin direction deviate from these axes. From the above discussion it should become clear that essentially the balance between exchange and magnetostatic energies is responsible for the existence of domains. Note, however, that the balance between exchange and crystalline anisotropies determines the width of the domain walls. The exchange interaction tries to keep the wall width large to allow only a slight variation of the angles of adjacent spins, while anisotropy favors abrupt changes of spin directions since it tries to lock the spins along the easy magnetization axes. This is the motivation behind a somewhat different mathematical description of the energy terms in a domain wall, in which the competing influences of exchange and anisotropy
energies within a wall are rewritten as a domain wall energy term E,. This term describes the energy needed to stabilize a wall of a certain type, thickness and length. Since within the domains the magnetization lies along a minimum anisotropy axis, domain formation is then given by the balance between magnetostatic and domain wall energies, the former favoring the creation of domains to save magnetostatic stray field energy, the latter trying to keep the domain wall length minimum. These arguments qualitatively explain why domains might form in a ferromagnet, be it in the bulk, at a surface or in a thin film. The quantitative argument, however, is much more intricate. First of all, the dimensions of the sample also play a role, as already seen in the demagnetization constant N in the shape anisotropy energy. For N = 0, the magnetostatic energy cannot be reduced further by the creation of domains. This explains why a thin film with in-plane magnetization is predicted to be a single domain. But even for more irregularly shaped samples, the situation might be more complex than can be inferred from the above discussion. Assume, for example, that the size of a sample is comparable to or even smaller than a typical domain wall thickness. Then a single-domain state will be preferred regardless of the large stray field energy, since no domain wall can form. A similar energy argument has been used by Kittel [5] for thin films of thickness d with the easy magnetization axis perpendicular to the film plane. He calculated the sum of magnetostatic and wall energy, E = E, + E,, for different domain arrangements. For the single-domain configuration one finds E,, a d, whereas for a multi-domain state without closure domains E mad I2 . Therefore, it was predicted that below a critical thickness a single domain prevails. Later Kittels pioneering work was generalized to the limit of vanishing film thickness by taking the magnetostatic interaction between the two film surfaces into account [6,71. Again a single-domain equilibrium state was predicted in zero applied field. All these calculations support the intuitive argument that for d -+ 0 the volume of the film vanishes, hence the stray field energy is negligible, and no domains need to form. Note, how-
163
Spin Analyzer /\
ever, that the area of a domain wall also vanishes for d + 0, and this complicates reality! Concurrently with the first experiments on ultrathin magnetic films, Yafet and Gyorgy [8] performed a detailed zero-temperature calculation which took the magnetic dipolar interaction between the single spins into account, distinguishing between the short-range and the long-range part of this magnetostatic interaction. They showed that under certain conditions a state with stripe domains has lower energy than a single-domain state in equilibrium. The film must be magnetized perpendicularly, and uniaxial anisotropy and magnetostatic energy should nearly cancel. Czech and Villain [9], by a discrete calculation on an lsing lattice with strong anisotropy, predicted a checkerboard domain structure to have an even lower energy. We stress again that the domain configuration is only controversial for perpendicularly magnetized films, since an infinitely extended in-plane magnetized film is stray-field-free even in the single-domain state. Therefore one of our main experimental efforts is devoted to understanding domain formation in out-of-plane magnetized films. We have investigated various systems, in order to also evaluate the influence of other critical experimental parameters, such as the quality of the substrate, the influence of temperature and magnetic field cycling. We were able to show that perpendicularly magnetized films generally decay into domains. In contrast to a 3D ferromagnet, the shape of these domains is very irregular. The size of the domains drastically varies from system to system, and currently is the subject of further studies. In-plane magnetized films, on the other hand, grow single-domain over huge areas, in agreement with theoretical predictions.
Secondary Electrons
An Fig. 1. Schematic drawing of the spin -,SEM principle. unpolarized focused electron beam scans along a ferromagnetic surface, and the spin polarization of the emitted secondary electrons is detected as a measure of the surface magnetization.
3. The technique of spin-SEM A technique for magnetic imaging must possess very high surface sensitivity to be able to probe the spin configuration in ultrathin films with thicknesses approaching a single monolayer. Furthermore, for the investigation of domain-wall profiles a high lateral resolution is advantageous.
Spin-polarized scanning electron microscopy with polarization analysis (spin-SEM [lo], or SEMPA [ll]) almost ideally fulfills the above requirements. Its principle was invented already some years before its first implementation [12] and is sketched in Fig. 1. A focused electron beam scans along the surface, thereby exciting a huge number of low energetic cascade electrons through multiple scattering processes. These cascade electrons serve as a fingerprint of the topographical, chemical, and magnetic state of the surface under investigation. The amount of emitted electrons is determined by the local topography of the sample surface (curvature, shadowing effects etc.) as well as by the chemistry through changes in the local work function and emission coefficients. These electrons are ejected into vacuum, and subsequently spin-analyzed. Their spin polarization is definedas P=(N, -N,)/(N, +N,),whereN? (N, ) is the number of electrons with spin parallel (antiparallel) to a chosen direction. The success of spin-polarized SEM relies on the facts that the yield of low energetic secondary electrons is high and that the spin polarization for these electrons is to a very good approximation proportional to the magnetization [13] within the topmost three to five monolayers (ML) of a surface [14,15]. Hence it is a very efficient technique to study ultrathin magnetic films of several ML thickness. In particular the direction of the polarization
164
of Magnetism and Magnetic Materials 129 (1994) 160-18.5
vector directly gives the orientation of the magnetization vector at such a sample surface. By measuring the direction of P, we thus image the magnetic domain distribution in these thin films. The experimental setup consists of a commercial scanning electron microscope that has been modified to achieve ultrahigh vacuum (base pressure < 1 X 10-i mbar), the prerequisite to maintaining clean, uncovered ultrathin magnetic films. The cold field-emission tip of the electron microscope delivers a finely focused beam of electrons, with a typically energy of 2 to 10 keV and a minimum beam diameter of 5 nm. When scanned along the sample, these primary electrons excite the secondary electron cascade. These electrons are extracted from the sample surface, energy filtered to select the lowest energetic electrons (typically 0 to 20 eV) and then spin-analyzed in a high-energy spherically symmetric Mott detector [16]. We are able to detect two polarization components simultaneously, one perpendicular to the sample surface, the other in-plane. The second in-plane component can be measured, if necessary, by a rotation of the sample around its surface normal. For detailed characterization most studies of magnetism in ultrathin films require the ability to vary the temperature. We have therefore built a special sample stage for our system, so that the samples can be cooled and/or annealed from T = 15 K up to T = 450 K, with only a minor compromise in resolution. Details on this stage will be published elsewhere [17]. In some cases it proved to be very convenient to characterize the films magnetically by a complementary technique. Thus an in situ magneto-optic Kerr effect [l&19] has been implemented, which allows small magnetic fields to be applied both in-plane and perpendicular to the sample surface. The ultrathin films are grown onto singlecrystal substrates by molecular beam epitaxy from effusion cells with integrated temperature control and water cooling. All films are evaporated onto substrates held at room temperature, except where stated otherwise, with a typical growth rate of 0.1 to 0.2 ML/min at a pressure of < 2 X lo- mbar. Chemical and structural characterization of the films is performed by Auger electron spec-
troscopy and low-energy electron diffraction (LEED). Growth of wedge-shaped films is also possible: A piezoelectric microslider [20] with a shutter mounted on top can be moved in front of the sample to block the evaporation beam. Step sizes as small as 10 nm are possibic. The entire system is equipped with a load-lock for easy transfer of a specimen from air to ultrahigh vacuum. Vibrationless pumping is achieved by ion-getter pumps supplemented by Ti sublimation.
4. Experimental thin films
observation
of domains
in ultra-
The theoretical predictions on domain formation in ultrathin films have been discussed above. Here I will present the corresponding experimental results that have so far been obtained. In particular films with their easy magnetization axis perpendicular to the film plane are compared with those magnetized in-plane in order to establish that in-plane magnetized films are indeed single domain over essentially the sample size and to find out whether this also holds for the controversial case with perpendicular easy axis. Various thin-film magnetic systems have been investigated. Moreover we have tried to judge the influence of other critical experimental variables, such as defects and roughness in substrate and film, temperature, or magnetic fields. The variation of these parameters might indeed be crucial, since it is not a priori obvious how the equilibrium domain state is reached. We find that in general perpendicularly magnetized films decay into domains, when they are grown without external magnetic field. The size of these domains varies drastically from system to system, and is currently the subject of further studies. In-plane magnetized films, on the other hand, behave as predicted theoretically: They grow single domain over huge areas. The first attempt to observe domains in ultrathin films concerned in-plane magnetized Fe/ Ag(100) layers [21]. Indeed, large in-plane domains were observed in a film of 3.4 ML thickness, in a 1 mm wide strip at the sample edge.
165
However, at smaller thicknesses no polarization could be detected, neither perpendicular nor inplane, in disagreement with earlier work in spinpolarized photoemission [22]. Meanwhile, evidence has accumulated that at room temperature thin Fe/ Ag(100) films are perpendicularly magnetized [23]. Recent results show that this is even true for films with a thickness of up to 6 ML [24]. Oepen et al. [25] were the first to succeed in imaging the magnetic state of in-plane magnetized 5-ML Co/Cu(lOO) films. He convincingly proved that these films are single domain on essentially the entire sample area of several millimeters. Only at the sample edge do closure domains form. By demagnetizing the film with an external magnetic ac field of decreasing amplitude, he was able to induce domains with magnetization directions along the four easy (110) axes. Both these early studies concerned in-plane magnetized films, the case that from a theoretical point of view is considered less interesting in terms of domain formation. We therefore started to investigate a class of magnetic systems that held greater promise for the observation of equilibrium domains, namely, thin epitaxial Co layers grown on (11 H-oriented face-centered cubic (fee) crystals. With standard magnetic characterization techniques that integrate over the sample size these systems display a striking similarity regardless of the substrate used, be it Au [26-291, Cu [291, Pd [30] or Pt [31-331. Below a certain thickness, they are magnetized perpendicular to the surface. Various models exist to account for this uniaxial surface anisotropy, e.g. one based on NCels crystalline surface anisotropy [341, one in which alloying at the interface occurs [35], or one that hinges on magnetoelastic effects, i.e. on strain induced by the lattice mismatch between film and substrate [36,371. The separation of these effects is quite a formidable task. A very recent study [38] on a qualitatively similar system, Ni/ Cm11 l), succeeded in sorting out these various contributions, owing to the very small lattice mismatch between Ni and Cu compared for example to that in Co and Au or Pt. In passing, we note that the intrinsic crystalline bulk anisotropy of hexagonal Co also contributes to an easy axis oriented perpendicularly to the surface. However, it is too
I 5w-l
Fig. 2. Magnetic domain image of a 3-ML hcp-Co/Au(lll) film, showing out-of-plane irregular domains with magnetization up (white) and down (black).
small to account for the out-of-plane magnetization without additional anisotropy terms. Thin, hexagonally-close-packed (hcp) Co films grown on Au(ll1) are the first example of this class of materials. The out-of-plane magnetization direction was established by hysteresis loop experiments [26]. With a spatially resolving technique such as spin-SEM we were able to verify this perpendicular easy magnetization axis directly by domain imaging without applying an external magnetic field [39]. Figure 2 shows the room-temperature magnetic domain image for a 3-ML Co/ Au(ll1) film, grown at room temperature with external magnetic fields shielded (H < 1 A/m). We found irregularly shaped domains of 2 to 5 Frn in size with the magnetization vector pointing along the surface normal. The simultaneously measured in-plane magnetization vanishes completely. The fact that the domains are fully out-of-plane proves that the uniaxial anisotropy is large enough so that no tilted magnetization direction occurs [26,40]. Figure 2 represents the first observation of an as-grown multi-domain structure. The question of course is whether the multi-domain state represents the equilibrium configuration or whether it is a metastable state that can be transformed to a single domain by e.g. temperature or field cycling. Note that the observation of domains is consistent with the finding that the very thin Co/ Au(ll1) films do not exhibit square hysteresis loops but instead have reduced remanence [41]. Thus a single-domain state is not the global energy minimum for these films. Possible reasons
166
why the multi-domain structure is the preferred arrangement are discussed in the following. According to Ref. [8], the relevant quantities are 6 = w/D (w = domain wall width, D = domain size) and f = K/E,, i.e. the ratio of total uniaxial anisotropy K to demagnetization energy. For domains to occur, f must be in the range 1 <f < 1.7 [42]. Such a small variation of f is predicted to cause 6 to vary by more than six orders of magnitude. This in turn leads to a dramatic domain growth, eventually to a size that is given by the macroscopic specimen diameter, i.e. several millimeters. From the domain images one can try to estimate 6 and hence also S. For our 3-ML film, D > 2 pm. We have not yet been able to resolve the domain wall width with our resolution of N 20 nm, and also the only technique with a higher resolution, differentialphase-contrast Lorentz microscopy [43], was unable to resolve this narrow transition [44] in the related system Co/Pt(lll). We therefore set w G 20 nm, which gives 6 < 0.01, and hence f > 1.04 [Sl. Thus one could expect finite domains. On the other hand, without resorting to the experimental determination of 6, f can be estimated from independently determined macroscopic magnetometry on Au/ Co/ Au(ll1) bilayers. We find f = 0.7 by taking the uniaxial anisotropy approximated by the surface anisotropy term KS = 0.53 mJ/m2 [45] divided by the film thickness of 3 ML, and the demagnetization energy containing the bulk magnetization M = 1440 kA/m. For such a low value of f, Ref. [8] predicts a single-domain state! Including the bulk anisotropy terms of hcp Co, K, = 412 kJ/m3 and K, = 143 kJ/m, modifies f to f = 1.1. This illustrates that small changes in the material parameters crucially affect f, and hence we cannot predict whether domains should occur. But we emphasize that varying the film thickness strongly alters the total anisotropy, and hence domains should eventually form at a certain thickness. Moreover, the direct applicability of Ref. [S] to our film is questionable for another reason. In Ref. [8] neither the influence of the finite temperature was taken into account nor the fact that both f and 6 are strongly thickness dependent. In fact Ref. 181 predicts a decreasing domain size with growing
film thickness, in striking disagreement with our observation of a linear increase of the domain size with the number of monolayers [46]. For the moment, I will postpone the observations related to a film thickness increase to Section 6. Here I want to elucidate some experimental facts about the Co/ Au(l11) system. A possible explanation for the existence and small size of the domains may be imperfections in the film, related to the growth particularities of Co/ Au(lll>. A recent scanning tunnelling microscopy (STM) study on the initial growth of Co/ Au(l11) showed polygonal bilayer island growth for the first two layers [47], owing to the 23 X 6 reconstruction of the Au(ll1) surface [48l, and a large lattice mismatch of 14% between hcp Co and fee Au. The typical island size is < 10 nm. For 3 ML, the films are contiguous, with a persisting threefold-symmetric arrangement of the Co hillocks, but no layer-by-layer growth in the classical sense is observed. Even at 7 ML, the granular structure resulting from the initial nucleation pattern is conserved. This explains the observed persistence of the multi-domain state up to larger thicknesses [39], where the domain pattern is essentially frozen-in during growth and then represents a metastable state, see Section 6. It is not entirely clear, however, whether the same argument holds for a film whose thickness is just at the onset of ferromagnetism. In such a model, each separate island arbitrarily takes one of the two possible magnetization directions during growth, and the resulting magnetization pattern is preserved in that patch after coalescence because of the large surface anisotropy. However, it is unlikely that an island with lateral dimensions of < 10 nm and a thickness of < 2 ML is ferromagnetic at room temperature. On the contrary, one expects that only a coalescence of these patches can overcome the percolation threshold and hence lead to magnetic long-range order. Moreover there is no straightforward connection between crystallographic and magnetic observations. The size of the patches is almost three orders of magnitude smaller than the observed domain size and the in-plane crystallographic symmetry is not reflected in the magnetic struc-
167
ture: The shape of the domains is completely irregular; neither three- nor sixfold symmetry is observed. In our opinion it is much more likely that structural imperfections influence the domain pattern by a local reduction of the domain wall energy. Thus, the multi-domain state can become the equilibrium state. It has been shown that the roughness of the films can be reduced by slight annealing and that correspondingly the surface anisotropy increases [27]. This change is also reflected in the domain pattern. As an example, Fig. 3 displays a 3-ML Co/ Au(ll1) film annealed to 620 K before a domain image was taken at room temperature. The domain size increases to lo-20 Fm. In no case, however, were we able to establish a singledomain film at zero applied field. This observation is in apparent contradiction to an investigation of magnetization reversal in Au/ Co/ Au(ll1) bilayers grown on glass [49]. However, their as-grown single-domain state might result from a persisting small external magnetic field [50]. In addition, the existence of a second interface influences both structure and magnetic behavior. It would be illuminating to directly compare the topography of annealed films and bilayers on glass with that of the as-grown single layer, so as to gain a deeper insight into the subtle relation between structural and magnetic properties. For these reasons we are currently investigating a second example of the same class of epitax-
Fig. 4. Magnetic domain film, showing out-of-plane (white) and down (black).
image of a 4-ML hcp-Co/Pt(lll) domains with magnetization
up
Fig. 3. Magnetic domain image of a 3-ML Co/Au(lll) film after annealing to T= 620 K, showing out-of-plane domains with magnetization up (white) and down (black). The measurement was performed at T = 300 K. Note the larger average domain size compared to that in Fig. 2.
ial systems, Co/Pt(lll). An extensive STM study revealed that the growth of Co on Pt(ll1) is even more delicate than on Au(ll1) and that it depends critically on substrate temperature and various other parameters [51]. By tuning these parameters properly, however, almost perfect layerby-layer growth can be obtained. This difference may partly be related to the smaller lattice mismatch of 10% for the Co/Pt system compared to the 14% of Co/Au. The difference in the strain at the interface influences both the growth of the Co film and the magnetic surface anisotropy, and may also determine the domain pattern. The perpendicular domains of a 4-ML hcp Co/Pt(lll) grown and measured at room temperature are displayed in Fig. 4. As for Co/ A&11), we observe highly irregular domain walls, but on a much larger scale: The domain size of the Co/ Pt(ll1) film is 20 to 50 km, i.e. one order of magnitude larger than in Co/ Au(ll1). A detailed study is under way that will hopefully help to resolve this puzzling observation on these two, at first sight rather similar systems. Ultimately, the aim is to connect structural and magnetic properties on a nanoscopic scale, e.g. by relating the occurrence and frequency of steps and defects to the observed magnetic properties such as domain size and surface anisotropy. In one favorable case we have already been able to show an intimate relation of topography and magnetism qualitatively on a microscopic scale, since it is possible to fabricate magnetic layers with completely different magnetic properties by changing the temperature of the substrate
168
during film growth. The system in question is fee Fe/Cu(lOO). Some years ago several authors compassionately discussed which growth mode would yield the films of best quality [52-551, but it is not my intention to resolve this matter here. Instead, I would like to show that, in contrast to Co/Au(lll) and Co/Pt(lll), there is no need to strain a film at the interface by as much as several percent in order to obtain a vastly different domain size. Minute changes in the growth mode of a film are sufficient for such an effect. Films of Fe were grown at two different substrate temperatures onto Cu(100). Although nominally of fee structure, there are slight tetragonal distortions perpendicular to the plane [56-581. Figure 5(a) shows a 2%ML fee-Fe/Cu(lOO) film evaporated at T = 300 K, while Fig. 5(b) displays a 3-ML fee-Fe/Cu(lOO) film grown at low substrate temperature, T = 90 K, and subsequently annealed to room temperature. The actual imaging experiments were done at T = 90 K and T = 240 K, respectively. Both films decay into a perpendicularly magnetized multi-domain state. In contrast to Co/Au(lll) or Co/Pt(lll), however,
the domain walls are pinned mostly at substrate imperfections visible in the topography [59]. The striking difference between these two films is the domain size. They differ by more than one order of magnitude! We do not yet understand the origin of this difference. LEED, the standard technique for investigating the surface structures, yields identical results for these two films: They both show the characteristic 4 x 1, 5 x 1 pattern [60-621 and have identical in-plane lattice constants. Macroscopic magnetic characterization by the magneto-optical Kerr effect is also not able to discern the differences observed with spin-SEM. Both films in Fig. 5 show rectangular hysteresis loops, with full perpendicular remanent magnetization. This example confirms that minute changes in the growth mode of ultrathin films play a key role for their microscopic magnetic behavior, be it through pinning at defects and dislocations in the film or substrate or through diffusion and intermixing at the interface. Several STM studies investigated the growth of fee-Fe/ Cu(100) films [63,64]. These experiments confirmed the nearly perfect layer-by-layer growth
(a)
I 1 Pm
100 pm
Fig. 5. (a) Domain pattern of a 2%ML fee-Fe/ Cu(100) film grown at T = 300 K and imaged at T = 90 K. Only the out-of-plane magnetization component is shown, the in-plane magnetization vanishes. (b) Domain pattern (left) and topography (right) of a 3.0-ML fee-Fe/Cu(lOO) film grown at T= 90 K, annealed to T= 300 K, and imaged at T= 240 K. Only the out-of-plane magnetization component is shown, the in-plane magnetization vanishes. Substrate imperfections visible in topography are seen. They serve as pinning sites of domains. Note the difference in scale between (a) and (b).
R. Allenspach/Journal of Magnetismand MagneticMaterials129 (1994) 160-185
169
once the first bilayer had formed at the interface. Although in general STM has no elemental sensitivity, Chambliss et al. [631 were able to detect a new type of intermixing at the interface for films grown at room temperature: Fe atoms replace Cu atoms in the top substrate layer. No such effect has been observed for films grown at low temperatures. This Fe-Cu intermixing has been confirmed by low-energy ion scattering, which owing to its elemental sensitivity can even quantitatively specify the amount of subsurface incorporation of Fe atoms [65]. Furthermore STM has shown that size and spacing of the Fe islands after deposition of a submonolayer coverage are N 8 nm for films grown at room temperature, compared to - 3 nm for films grown at low temperature [631. It is not known how this difference affects the thickness range of the films shown in Fig. 5, but I find it difficult to imagine that the smaller patches in the films grown at low temperature evolve into flatter, more perfect layers. On the other hand, if the domain size is assumed to be mainly given by structural defects one fails to even qualitatively understand how smaller patches correlate with the observation of much larger domains! Hence the correlation between topography and magnetism is much subtler. Minute alterations in the structure of interface and film are responsible for the drastic difference observed in our microscopic magnetic characterization, possibly via the surface or interface anisotropy, the ultimate connection between crystal lattice and magnetic moment on the atomic scale. Theoretically it is clear that the domain size in a bulk 3D specimen resists a reasonable quantification, because the shape of the specimen, its magnetic history as well as structural imperfections all contribute to that size. This situation might be easier to handle in ultrathin 2D films, where at least the shape is fixed. A very recent calculation based on Refs. [7-91 has yielded a formula for the domain size in the ultrathin-film limit [66]. Its remarkable result is that the ground-state domain size is determined only by the intrinsic magnetic and structural properties of the film, namely, by magnetization, anisotropy as well as lattice constant and film thickness. In addition, Kaplan and Gehring were the first who
were able to explain the linear dependence of the domain size on film thickness we had observed in Co/ Au(lll> films with perpendicular magnetization [46], which was in striking disagreement with the calculations of Refs. [51 and [81. There was also an earlier attempt to explain the linear dependence [67], unfortunately with a totally unphysical model. I strongly emphasize the technological importance of such an explicit relation between material properties and magnetic domain size. If such a connection is a universal feature of ultrathin films, this might ultimately determine the minimum size a bit must have to be thermodynamically stable. Note that Kaplan and Gehring [661 also found an argument why the domain shape is completely irregular. In their reanalysis of Refs. [S] and [91, they found only a minor energy difference between stripe and checkerboard domains. If these patterns have similar energies despite the difference in the total domain wall lengths of a factor of two, then any irregular pattern should also have essentially the same total energy. I conclude this section by summarizing the differences between in-plane and out-of-plane magnetized films. From time to time, the experimentalists luckily discover systems in which epitaxial layer-by-layer growth can be achieved. But there are even more exciting coincidences: On Cu(lOO), both fee-Fe and fee-Co films can be grown with high perfection, and the magnetization direction is out-of-plane for Fe/ Cu(100) and in-plane for Co/Cu(lOO), as discussed above. We have therefore reinvestigated the in-plane Co/ Cu(100) films that had been characterized by Oepen [251, with the aim of directly comparing Fe and Co films grown on the identical substrate single crystal and under identical external conditions. Our results confirm the single-domain state of ultrathin Co/Cu(lOO) films 1681.In particular, if we compare Fe/Cu(lOO) with Co/Cu(lOO), we find that the same defects in the substrate are sufficient to serve as pinning centers for out-ofplane reversed domains but not for in-plane domains. This might be considered a trivial fact, especially in terms of the classical case of a 1D chain consisting of magnetic dipoles. In a chain where
170
of Magnerism and Magnetic Materials 129 (1994) 160-185
all dipoles are oriented perpendicularly to the chain, the local demagnetizing field will favor a disturbance that tends to reverse a single dipole. On the other hand, in a chain where the dipoles point along the chain axis, the local field helps to align a disturbed dipole. This simple picture is the 1D illustration of shape anisotropy, as discussed earlier. However it does not contain an explicit exchange interaction and hence does not describe domain formation completely. In my opinion it.nPvertheless proves that the small magnetostatic energy is responsible for the distinct difference between in-plane and out-of-plane magnetized films - in contrast to the theoretical predictions prevailing until recently, which were based on the argument that the film volume is virtually negligible. Our experiments furthermore show that Yafets criterion [S] for the formation of domains in perpendicularly magnetized films is too restrictive to agree with the observations in completely different systems like Co/ A~(11 11, Co/ Pt(ll1) or Fe/ Cu(100). The stripe-domain phase, however, predicted to occur where the uniaxial anisotropy and the shape anisotropy almost compensate, has indeed been identified very recently [69]. I will come back to this topic in a somewhat different context in detail in Section 6.
5. Two-dimensional magnetism the Curie temperature
below and above
One of the most fundamental quantities in magnetism is the spontaneous magnetization Ms, and especially its temperature dependence M,(T). On the one hand Ms directly reveals the quantum mechanical nature of the exchange interaction, on the other hand it is the order parameter that characterizes the ferromagnetic to paramagnetic phase transition. The Curie temis the transition temperature at perature, T,, which the spontaneous magnetization vanishes and thus designates the point where the concepts of quantum mechanics and the modern theories on phase transitions come together. In his seminal work Weiss [l] explained the temperature dependence of the spontaneous magnetization by inventing a large molecular field
that is strong enough to align the magnetic moments against thermal fluctuations. If modified by this molecular field the Langevin theory [70] for a paramagnet successfully explains the temperature dependence of the spontaneous magnetization. Specifically it predicts a phase transition at a certain temperature where the tendency for the moments to remain parallel is totally compensated by the thermal energy. Given the simplicity of the model, the phase transition in this mean field theory is astonishingly accurate. In particular the model reveals universal properties predicted for a phase transition by more advanced scaling theories. Although the postulate of such a molecular field was very successful, its origin was completely mysterious at the time. Quantum mechanics finally explained this huge field by the exchange interaction between the electron spins. Despite the success of mean-field theory, discrepancies to experiment became obvious even in bulk 3D ferromagnets [71]. At low temperatures, spin-wave excitations [72,73] occur, and in close proximity to Curie temperature, spin-spin correlations are important. Theories on phase transitions quite generally predict that such correlations and fluctuations become more pronounced in low-dimensional systems. Thus one can expect a wealth of unusual phenomena in ultrathin films, which probably are the most perfect realization of a 2D system to date. The most extensively studied features of a phase transition are the various critical exponents. The universality hypothesis describes the second-order phase transition from a ferromagnet to a paramagnet by M,(T)/M,(T = 0) = (1 - T/Tcjp, wh ere /3 is the critical exponent describing the phase transition of the spontaneous magnetization. The mean-field value /? = 1 is modified depending on the dimensionality of the system and the degrees of freedom of the spin. Hence p is different in 3D bulk, at a surface or in an ultrathin film. In 2D, theory predicts B to be exactly 0.125 for an Ising model, but near 0.23 for a finite-size XY model [74]. Numerous investigations for ultrathin films exist that either confirm or refute the theoretical value of p. Usually spatially averaging techniques have been applied to determine both /3 and Tc
of Magnetism and Magnetic Materials I29 (1994) 160-185
[75-791. Implicitly this approach assumes that the microscopic and macroscopic magnetic properties are identical at all temperatures between 0 G T G T,. Such a conjecture, however, is only justified as long as the domain configuration and hysteresis loops do not change with temperature. The aim of our study was not to determine the critical exponents in an ultrathin film once more, but rather to investigate the magnetic response of an ultrathin film in the vicinity of T, in detail. We combined spin-SEM and the magneto-optic Kerr effect to establish the correspondence between the microscopic and macroscopic quantities, i.e. spontaneous magnetization and remanent magnetization. As a model system we investigated l-ML fee-Co/Cu(lOO) films magnetized in-plane [80]. These films grow nearly layer-by-layer, as recently verified in a detailed STM study [Sl]. We followed the temperature dependence of these quantities between 0.4 and 1.2 T/T, by measuring hysteresis loops with the Kerr effect and domain images with spin-SEM. From Kerr hysteresis loops we have determined the remanent magnetization M, = M (H = O), and the magnetization in an applied field of 8 kA/m, M, = M(H = 8 kA/m), see Fig. 6(a). We have observed the following remarkable features as a function of temperature. The remanent magnetization vanishes sharply at a well-defined temperature that tentatively is identified with the Curie temperature Tc of the film. At temperatures below 0.9 T/T,, M, and M, coincide, and the hysteresis loops are square. However, for T > 0.9T,, the loop shape starts to deteriorate, and M,<M,. In particular, for T> T,, the hysteresis loop is replaced by a curve reminiscent of a typical Langevin function for a ferromagnet [821, with M, = 0, while M, attains sizable values up to T > l.lTc. This observation that a large magnetic response is induced above T, with very moderate fields of the order of several kA/m is an unusual observation in 3D magnetism, except if superparamagnetic relaxation is present. We should then identify our T, as a blocking temperature Ts, and the Curie temperature itself would be higher. I would like to emphasize that it is quite important to identify the Curie temperature ex-
(4
/ I
0.4
0.5
0.6
0.7
0.8 T/T,
0.9
1.0
1.1
1.2
Temperature dependence of the Kerr intensity for a * l-ML Co/Cu(lOO) film at an applied field of H = 0 (M,, solid circles) and in an applied field of H = 8 kA/m CM,, open circles). Two hysteresis loops are shown as insets at temperatures T/ Tc = 0.90 and T/ Tc = 1.03, respectively. The square loop is representative for all temperatures T d 0.90 Tc. (b) Temperature dependence of the spin polarization (or spontaneous magnetization Ms) deduced from scans across oppositely magnetized domains. No external field has been applied. Solid circles: in-plane magnetization; open circles: out-of-plane magnetization (which vanishes completely). Note the perfect agreement with Ma as determined from the Kerr effect, Curie temperature Tc = 289.5 K.
actly, and to pin down the influence of an external field. For an experimental determination of critical exponents by a fit, the Curie temperature must be known. Various different approaches exist for this. Either T, is defined as the vanishing of the remanence 1761, of the saturation [78,79,83] or of a saturation extrapolated to H - 0 [84]. T, has also been determined by recording the divergence of the susceptibility [85]. From a
172
of Mugnetism and Magnetic Materials 129 (1994) 160-185 decoupled enough to show superparamagnetic phenomena, like e.g. relaxation. Indeed with Mijssbauer spectroscopy it was possible to show that 2-ML Fe/ Ag(l11) superlattices 1903 and 2-ML Fe/ Au(ll1) films [84] have superparamagnetic characteristics. Recent structural studies on Fe/Ag(lll) by LEED [91] and on Fe/Au(lll) by STM [92,93] confirm a growth in 3D islands, or nonideal layers. We are thus confronted with two possible interpretations of the Kerr data in Fig. 6. Either the strong response of M,, above T, is caused by intrinsic properties of a 2D ferromagnet, i.e. strong thermal fluctuations, or the film is a superparamagnet. In the latter case, one would identify the temperature at which M, vanishes with a blocking temperature rather than with a Curie temperature. One therefore has to establish whether the T,. defined in Fig. 6 is indeed the Curie temperature of the Co film or, equivalently, whether the spontaneous magnetization M, is identical to the remanent magnetization M, in the entire temperature range, and especially near Tc. To distinguish the scenarios discussed above the question of the nature of the phase transition has been addressed experimentally. We used our spin-SEM precisely to determine M, CT). The spin polarization P of the secondary electrons emitted from a single domain locally measures a quantity proportional to the spontaneous magnetization, with a lateral resolution of - 20 nm. We made sure that at each temperature we determined P or M, within one domain. The results are displayed in Fig. 6(b). There is complete agreement between M,(T) and M,(T), i.e. between the macroscopic and microscopic quantities. Moreover, we observe neither a change in domain pattern up to 1 K below T, [81] nor a relaxation on our experimental time scale of several minutes. Furthermore the disappearance of M, occurs in a much sharper temperature interval than expected for a superparamagnet [84]. Hence we rule out superparamagnetism as the cause for the vanishing of M, and indeed identify T,. as the Curie temperature of the system. Therefore, in Fig. 6 we interpret the strong response of the magnetization M,, to a moderate applied field above T,- as an intrinsic property of
theoretical point of view, it is clear that the spontaneous magnetization is the order parameter for the phase transition, as discussed above. How the remanent magnetization behaves when T approaches T, is much less well understood. In ultrathin films it has generally been taken for granted that the hysteresis loops are square up to T,, and hence M, = M, for all temperatures. Figure 6(a) proves that exactly in the temperature range where the phase transition occurs this hypothesis is not fulfilled. What is the reason for our observation? One could argue that a loss of M, concurrently with a sizeable M, could be caused by the occurrence of opposite static magnetic domains, which tend to reduce the macroscopic remanence. This in turn would imply that the Curie temperature in our films is much higher than what we defined as T,. Indeed a theoretical argument exists suggesting that domain formation is a critical process 1861. I am, however, not aware of an experimental verification of such a phase transition. In fact an observation by Gradmann et al. [87] in Fe/W(llO) films refutes such a criticality of domain formation. They magnetometrically detected a sizeable magnetization in a small applied field at temperatures for which the hyperfine splitting in Miissbauer spectroscopy has already collapsed to zero. Miissbauer spectroscopy as an ideal local probe is insensitive to the presence of magnetic domains. Thus the disappearance of the hyperfine field cannot be related to a change in the domain pattern. A more likely explanation of our Kerr data would be a superparamagnetic response. In a superparamagnet [88] small entities (grains or islands) behave as paramagnetic particles with a huge magnetic moment. Above a certain temperature, the blocking temperature T,, thermal excitations make the magnetization of the particles fluctuate rapidly, and hence the averaged quantity M, = 0. Below T,, the thermal energy is smaller than the anisotropy energy in the particle, the magnetization direction is fixed, and therefore M, > 0. Small 3D particles generally exhibit these properties and have been studied widely [89]. In 2D epitaxial systems one can imagine that films growing in an island-type fashion might be
173
a truly 2D ferromagnet: There are large thermal fluctuations of spin blocks [941. These blocks act like a giant spin. They easily orient along a small field, and induce the appreciable magnetization at and above Tc. It is possible to estimate the field-induced magnetization in a mean-field approach [80]. We find M,(T,)/M,(T = 0) = 0.5, in qualitative agreement with the observations, see Fig. 6. The same calculation yields M,(T,)/M,(T = 0) = 0.01 for a bulk ferromagnet. This is a typical value for a 3D magnet, and indeed such a response was experimentally observed already a long time ago [95]. This estimate again nicely illustrates the fundamentally different role the spin fluctuations play in 2D and 3D magnets. Note that the physical concept of the spin blocks is reminiscent of superparamagnetism, but subtle differences exist. First, the spin blocks - in contrast to superparamagnetic grains - are not pinned at defects on the sample, but are allowed to move freely. Second, the size of the spin blocks strongly varies with temperature owing to the T-dependence of the correlation function [94]. A direct consequence is that the magnetization loop M = M(H) above T, - although similar to a Langevin function - should rather be described by a linear curve MN H at low fields, followed by a logarithmic dependence A4 N In H at intermediate fields [96,97]. I expect that our observations in this Co/ Cu(100) film are a common phenomenon in 2D ferromagnets. Indeed experimental evidence exists that Fe/Ag(lOO) also shows large spin-spin correlations and that the films behave as 2D Heisenberg ferromagnets [98]. In this case it was possible to disprove superparamagnetic behavior by measuring the transmission Mossbauer spectra for magnetizing along easy as well as hard magnetization axes. The early Fe/W(llO) data mentioned above [871 were also a clear sign of a truly 2D ferromagnet with strong fluctuations above T,. Very recently this has been verified in a study on the temperature dependence of the magnetic anisotropies [99]. I argue that also epitaxial Fe/W(lOO) [lOOI essentially shows a similar behavior in terms of a proper 2D ferromagnet. Perfect epitaxy, however, is not mandatory. Even in sputtered ultrathin permalloy Ni,,Fe,, films,
the relevant characteristics of a 2D ferromagnet have been identified [73]. The hysteresis loop shapes display analogous behavior with temperature as our perfect, epitaxial Co/Cu(lOO) film does. Another sputtered 2D magnet with very interesting magnetic properties has recently been reported [97]; namely diluted Pd(1.2 at% Fe) films sandwiched between thick Pd layers. On the one hand this system shows pure 2D Heisenberg behavior for T > 6 K, while on the other hand, for T < 5 K, relaxation phenomena reminiscent for a superparamagnet set in, specifically, the remanent magnetization decays with time. I do not understand these seemingly conflicting results. They might be related to a crossover from a Heisenberg model to a different universality class, such as for example an XY model. Such crossovers have been found in layered magnets earlier [lOl1031. We have learned that great caution is needed when performing experiments to determine reliable critical exponents or Curie temperatures in 2D ferromagnets [104]. It is essential that external fields are completely shielded because of the giant response of the magnetization even to small fields. This effect is expected to be more pronounced in in-plane magnetized films, but may also be present with perpendicular magnetization. Therefore I would like to emphasize that the interpretation of the results for Co/Cu(lll) [79] should be reexamined in view of the equivalence of remanence and spontaneous rather than saturation magnetization. Recently it was brought to our attention that the nuclear magnet He on graphite [105,106] behaves quite analogously to the eZectronic ferromagnet Co/Cu(lOO), and might be a even more perfect realization of a 2D Heisenberg magnetic system [107,1081. The beauty of physics is demonstrated here at its best: Natures universality among seemingly completely different fields of physics. 6. The role of anisotropies: tion switching
Magnetization direc-
One of the most fascinating features in ultrathin films is that 2D ferromagnetism is stabilized
114
exclusively by magnetic anisotropies. The Mermin-Wagner theorem [109] proves that an isotropic 2D Heisenberg magnet does not exhibit long-range order at finite temperature. The ultrathin films discussed in Section 5 can be considered as an approximate realization of an isotropic 2D magnet, but only above T,. Below T,, an almost isotropic system can be imagined in which different terms of the total anisotropy energy cancel. Several anisotropy contributions exist. The most prominent ones have already been discussed in Sections 2 and 4. The uniaxial surface or interface anisotropy is responsible for the perpendicular magnetization direction observed in many ultrathin films. The shape anisotropy, on the other hand, favors a magnetization parallel to the plane and, moreover, is the relevant energy term for domain formation. It is therefore tempting to investigate the situation in which these anisotropies nearly compensate. In a phenomenological approach the competition between surface and shape anisotropies determines the magnetization direction. If the balance between these quantities is changed, a change of the magnetization direction can occur, from an orientation perpendicular to the film surface towards the plane. Remarkably this subtle balance is experimentally approachable since it hinges on different parameters such as temperature, film thickness or magnetic field. A more fundamental approach has to connect phenomenology with the intrinsic properties of 2D systems. Recent theoretical studies emphasize the importance of thermodynamic fluctuations [llO-1131. The magnetization reorientation must be considered as a phase transition below T,, and it is essentially the result of the entropy of disorder prevailing in the free energy [llO,lll]. The phase transition was calculated to occur as a discontinuous switching from a perpendicular to an in-plane magnetization in a single monolayer [ill]. Later it was shown that the directional entropy can be responsible for a continuous transition [112]. A further study [113] predicts that in the vicinity of the magnetization-switching temperature, T,, a finite temperature interval Ai?, exists in which long-range ferromagnetic order is absent. It has been emphasized [112] that calcula-
tions should also be performed with the film thickness as a parameter and that in-plane anisotropies at temperatures above T, should be taken into account. Such a reorientation transition certainly does not only occur in a film of fixed thickness with varying temperature. It is equally likely that at a fixed temperature any film magnetized normal to the surface will change its magnetization direction at a certain thickness d, owing to the predominance of shape anisotropy over crystalline surface anisotropy. Existing theories restrict themselves to the first case; experimentally, however, both situations have been investigated. An easily accessible quantity for monitoring the transition from perpendicular to in-plane magnetization is the remanent magnetization vector M,. The first case, M,(T = const, d), has for instance been measured for Co/ Au(ll1) [26], Fe/ Cu(100) [231 and Fe/ Ag(100) [23,24,114]. The second case, M,(T, d = const), is realized in Fe/Cu(lOO) [60,1151, Fe/Ag(lOO) [23,24], and CuXCo,_,/Cu(lOO) [116]. The transition has also been observed in Fe/ Ag(100) films [1171 and superlattices [1181 with other techniques. In Mossbauer spectra a temperatureand thickness-dependent change in the peak ratios has been observed which is a clear signature of a magnetization direction change from perpendicular to parallel. Quite astonishing is that among these systems completely different behavior is observed. With increasing film thickness, the perpendicular remanent magnetization in Co/ Au(ll1) decays smoothly near the switching thickness d,, and the in-plane remanent magnetization sets in correspondingly, thus compensating the reduction of the perpendicular component. In Fe/Cu(lOO), however, the perpendicular remanence decays and vanishes completely before the in-plane remanence sets in at higher thickness. We have therefore performed experiments in these two systems, hcp Co/Au(lll) and fee Fe/ Cu(lOO), to monitor the magnetization direction switching by a spatially resolving technique. In the case of Co/Au(lll) we were interested in whether this transition is a smooth rotation or has two discrete states, vertical and in-plane. In
175
the case of Fe/Cu(lOO) we wanted to know whether an approximate isotropic system is realized, i.e. whether the loss of remanence is accompanied by a loss of spontaneous magnetization. Therefore the central question for both systems is how such a reorientation transition influences the domain pattern in a 2D ferromagnet. 6.1. Co /AdIll) films
3 ML
4 ML
I first summarize the Co/ A~(1111 results [391. A thin Co film has been grown and its thickness subsequently increased in well-defined intervals. At each thickness, the domain pattern of the identical 20 km by 20 pm area on the sample has been imaged. The results of such an experiment are shown in Fig. 7 for the thickness range 3 ML G d G 6 ML, each with the out-of-plane and one in-plane component. Below 2 ML we find no magnetization at room temperature, since T, < 300 K. At about 2 ML, tiny domains of N 500 nm diameter are identified (not shown). At 3 ML, an out-of-plane domain pattern similar to that in Fig. 2 can be discerned. One additional layer forces the small domains to coalesce into larger domains, and some faint contrast is now visible in the in-plane magnetization component. The switching thickness d, is reached after evaporation of an additional 0.5 ML. At 5 ML the film magnetization has essentially turned into plane. Above d 2 6 ML, no perpendicular component is left within the statistical error of the experiment, but an in-plane domain configuration is found. This does not contradict our statement in Section 4 that a single-domain state is the ground state of an in-plane magnetized film. By applying an external magnetic field pulse, a homogeneously magnetized state is reached [39], and the hysteresis loop becomes rectangular. The observed multi-domain in-plane pattern is therefore governed by the existence of domains in the perpendicularly magnetized thinner film and, hence, by a nonequilibrium magnetic state frozen-in during growth. By simple visual inspection of these images we reach our qualitative main conclusion. The reorientation occurs gradually by a rotation of the magnetization direction. More quantitatively, the
4.5 ML
5 ML
6 ML
I
10
pm
Fig. 7. Magnetic domain images for Co/Au(lll) thin films showing the evolution of domain size and switching behavior vs. film thickness. The pictures were taken at identical positions on the sample between evaporation of additional layers. For each thickness, the image on the left gives the in-plane magnetization component along the length of the page CM,, ) and that on the right the out-of-plane component CM, 1. Gray scales from black to white indicate the magnitude of the magnetization component along the measured axis. Note the coalescence of the domains on proceeding from 3 to 4 ML thickness.
angle 0 between surface normal and magnetization can be extracted from Fig. 7 for each individual point and averaged, see Fig. 8. We find a continuous rotation from out-of-plane to in-plane magnetization in a finite thickness interval of 2 ML. In particular, a tilted magnetization direction is observed at a switching thickness ds = 4.5
176
ing on the actual 6 = o, 0 = 90, and sin20 = (~/2
numbers,
the
solutions
are
X M2 -K, -Kv'd)/W,).
I
1 .o
I
7.0
2.0
3.0 4.0 Film Thickness
6.0 5.0 (ML)
Fig. 8. Angular dependence of the magnetization direction vs. thickness derived from point-by-point analysis of Fig. 7. 0 is the angle between surface normal and magnetization. The solid line represents the best fit for sin*0 = (IL /2 x M2 - K, - K,/d)/(2K,), with the Co bulk values M= 1440 kA/m, K, = 412 kJ/m3, K, = 143 kJ/m3, and the fit parameter KS = 0.62 mJ/m2. The dashed lines give the fits for K, = 0.57 mJ/m2 and K, = 0.67 mJ/m, respectively, for comparison.
ML, with no preferred direction of the in-plane components. A simple energy consideration suffices as a starting point to understand the continuous magnetization reorientation with the tilted configuration during switching: It is caused by a different angular dependence of the demagnetization energy and second hexagonal anisotropy constant K,. As described in Section 2, the total energy within a single domain consists of exchange energy, magnetocrystalline anisotropy energy, and demagnetization energy, E = E, + EA + E,. During reorientation the exchange term does not vary and can therefore be neglected. The crystalline anisotropy of the hexagonal Co film is EA = K, sin20 + K, sin40 + KS/d x sin20, where K,, K, are the uniaxial anisotropy constants of bulk Co and the last term contains the surface anisotropy of the ultrathin film [37,45,1191. The demagnetization energy within a single domain, E, = po/ 2 x NM2, is rewritten to include the angle between surface normal and magnetization M explicitly: E, = p0/2 X M2 cos28. Minimization of the total energy with respect to 8 determines the equilibrium values for 8. Depend-
It is this third solution that leads to a tilted magnetization direction, since it contains a nontrivial f3 = B(d) dependence. From our data one can thus determine the uniaxial surface or interface anisotropy K, by taking the Co bulk values for M = 1440 kA/m, K, = 412 kJ/m, and K, = 143 kJ/m. The result of the fit in Fig. 8 is K, = 0.62 & 0.05 mJ/m2, in good agreement with the 0.53 mJ/m2 found earlier [45]. We emphasize that the phenomenological quantity K, allows no insight into the physical origin of the uniaxial anisotropy. The possible contributions to K, and the difficulty of distinguishing them unequivocally have already briefly been mentioned in Section 4. For Co/ Au(ll1) with the large lattice mismatch of 14% strain might be the leading term in the surface anisotropy. In fact, for such a lattice mismatch strain already induces misfit dislocations in the first bilayer at the interface [45,471. We do not think that alloying plays a role in Co/Au films, because Co and Au are virtually immiscible. Since a large perpendicular surface anisotropy is quite a common phenomenon in Co films grown on fcc(l 111 substrates, we - and also others [37] are tempted to expect a common cause for such a magnetic behavior, most likely strain. I therefore believe that the interpretation of K, in terms of an alloy phase at the interface in Co/Pt(lll) multilayers [35] should be reexamined, in particular also in view of the recent results of Ref. [381. Nevertheless, experiments with unstrained hexagonal Co films would be highly desirable. I would like to add a final remark about the nature of the magnetization reorientation. A continuous rotation during a finite thickness interval is not necessarily the only solution of the energy minimization. The material-dependent quantities (magnetization, anisotropies) determine the specific behavior. For Co/ A~(11 l), it is the balance between magnetization on the one hand and K, + KS/d on the other hand that determines the switching thickness d,, owing to the competition between demagnetization energy and anisotropy
of Magnetism and Magnetic Materials 129
(1994)
160-185
177
energy. The existence of a finite K, with a different angular dependence than M and K, or K,, finally, determines the finite transition width responsible for the continuous rotation. To my knowledge, this is one of the very few examples which clearly shows the importance of the second-order anisotropy constant, a constant that usually can be neglected in most calculations and experiments. Note that the simple energy consideration did not include a higher-order anisotropy term which would take in-plane anisotropies into account. This omission is justified by the experimental observation that no sixfold symmetry in the perpendicular domain pattern exists (see Section 4) and that no orientation is preferred during the rotation of the magnetization direction from perpendicular to in-plane. 6.2. Fe / Cu (100) films We now turn to the investigation of the magnetization direction reorientation in fee Fe/ Cu(100). For this purpose we have grown an Fe wedge on top of the Cu(100) single crystal. After 30 years of obscurity [120] wedge-shaped samples became fashionable two years ago [121,122]. The appealing feature of wedge samples is that with a spatially resolving technique such as spinSEM it is possible to map out a complete vectorial thickness dependence M (T = const, d) in one experiment. Moreover, this eliminates most experimental uncertainties between different films, such as for example the relative film thickness or different growth conditions. We made the Fe wedges following the two recipes generally used: In the first case, films are grown onto Cu(100) at or near room temperature, in our case T = 260-280 K. In the second case, films are evaporated onto the substrate held at low temperature, T = 90-100 K, which is known to produce sharp interfaces between substrate and film [54]. Subsequently the films are annealed to T = 300 K. These two approaches yield epitaxial fee-Fe/ Cu(100) films with completely different magnetic properties. In Section 4 we have seen that for example the size of the perpendicular domains in the as-grown films varies by more than one order of magnitude. The magnetization direction
switching from perpendicular into the plane is also completely different. Films grown at room temperature exhibit diverse - and strongly disputed - magnetic properties in a thickness range between 5 and 12 ML [52,62,123-1251 that most likely are connected to a change in structure [57,62]. It has been shown that only a part of the film is still ferromagnetic [1231, and that the Curie temperature is reduced [123,126]. Spatially resolved experiments on these films can provide new insights regarding the location of the magnetically active layer. I will not discuss these results here, since the aim of this section is to gain an understanding of the pure magnetization direction reorientation, which is not disturbed by structural changes whose magnetic influence is hard to quantify a priori. Fe/Cu(lOO) films fabricated according to the second recipe (growth near T = 100 K, annealing to T = 300 K> fulfill these requirements. Pappas et al. established the magnetic reorientation in Fe/ Cu(100) films of exactly this growth type in their spinpolarized secondary electron spectroscopy studies [23,60,115]. We therefore can directly compare the spatially resolved experiments with their integrating technique. In fact their findings stimulated our own investigation of the magnetization reorientation because of several provocative observations. Hence, I first summarize the main results of Refs. [23,60,115]. These authors monitor the remanent magnetization vector M, during the transition from perpendicular to in-plane magnetization by measuring the spin polarization of the secondary electron cascade emitted from an area of typically 1 mm2. All three directions of M, are measured sequentially after applying external field pulses. A film thickness range of 5 ML <d, G 6.5 ML is found, in which a reversible magnetization direction switching occurs at temperatures between 300 K > T, > 140 K. The most intriguing finding is that M, disappears along all three spatial directions at the film thickness d, where switching occurs. Even more striking is that M, vanishes not only at the switching temperature T, but also in an extended temperature interval of AT, = 30 K. Although a finite AT, can be understood in the framework of Ref. [113], the theoret-
178
ofMagnetism and Magnetic Materials 129 (1994) 160-185
ical explanation cannot be considered satisfactory: The experimental value for AT, is more than one order of magnitude larger than the calculated one! We can expect that A.7; is intrinsic to the 2D nature of these films. It has been speculated that a paramagnetic intermediate state might exist, since the surface anisotropy and the short-range part of the shape anisotropy cancel at T,. The symmetry arguments of Ref. [109] - although not strictly fulfilled because of the longrange part of shape anisotropy - might then become more relevant. Another likely possibility is that domains form during reorientation, as conceivable in the spirit of Ref. [S]. Note that with these experimental results even an antiferromagnetic state cannot be ruled out [23]. Furthermore it has been found that M, is considerably reduced in films whose thickness d > d,. This decrease has been ascribed to a loss of magnetic moment from 2.2~~ to 1.4~~ in fee Fe/Cu(lOO) [231. We note that such a reduction has not been observed in bee Fe/ Ag(100). Because of this dissimilarity one could argue that the reduced M, in Fe/Cu(lOO) is a consequence of a structural change in the film at - 6 ML, similar to the one observed for films grown at room temperature [56]. We were able to resolve these fundamental aspects of the magnetization orientation change in fee Fe/ Cut 100) [691 by locally investigating the switching behavior with increasing film thickness and temperature. These spinSEM results will be reported in the following. Figure 9 shows a wedge sample with 0 G d G 7 ML fee Fe/Cu(lOO) grown at T = 90 K and subsequently annealed to 300 K. The actual imaging experiment takes place at T = 200 K. For d < 2 ML, no magnetization is found, neither perpendicular nor parallel to the sample surface, hence T, < 200 K. Within 2 ML < d < 6 ML, perpendicular magnetization prevails. Above 6 ML, in-plane magnetization is found, with the magnetization lying along the [OOll direction. Note that the out-of-plane magnetized part of the wedge has been driven to a single domain by applying a field pulse of H = 4 kA/m to facilitate the following analysis. A line scan along the wedge directly gives the spontaneous magnetization M,(d). Fig-
I 7ML
lOOl1
I I
20 pm
Fig. 9. Magnetic domain images of a 0 < d < 7 ML Fe wedge atop the Cu(100) substrate, grown at T = 90 K and annealec to T = 300 K. A side view of the position of the wedge relative to the images is shown at the top. Upper panel: magnetization component perpendicular to the surface; lower panel: magnetization component parallel to the surface along the (0011 direction. Gray scale varies from spin polarization P = - 30% (black) to P = +30% (white). The film is nonmagnetic for d < 2 ML, perpendicularly magnetized for 2 ML < d < 6 ML, and in-plane magnetized for d > 6 ML. Small in-plane domains form at crossover thickness. Measurement temperature: T = 200 K.
ure 10 displays both the perpendicular and parallel magnetization components along the [OOl] direction, MS I and MS,, , as well as the calculated magnitude of the spontaneous magnetization MS = {(MS L>~ + (MS ll)2)2. Two features are noteworthy: First the spontaneous magnetization is not reduced for the in-plane thickness range (d > d,), but increases as expected for a thickness-independent magnetic moment, owing to the small escape depth of secondary electrons. The earlier discovery of a reduced remanent magnetization must therefore be ascribed to the inability to establish a single-domain state over the probed area of 1 mm2. Recent calculations have predicted that the magnetic moment may depend on the magnetization direction, since spin-orbit coupling might no longer be quenched in ultrathin films [1281. This effect is predicted to be small, of
179
10 4-_,_: 0
I
1 2
/
3 4 5 (ML) 6 7 Film Thickness
Fig. 10. Line scan across domain image of the wedge sample in Fig. 9 displaying the (a) perpendicular (MS L, solid circles) or spontaand parallel CM, ,,, open circles) spin polarization neous magnetization vs. Fe film thickness, and (b) the calculated total spontaneous magnetization MS = {(MS I 1 + (MS ,,)2)1/2 vs. film thickness, as calculated from (a). The spontaneous magnetization increases with Fe film thickness as expected for a constant magnetic moment, owing to the escape depth of secondary electrons of w 4 ML. From [127].
the order of a few percent. One cannot expect to observe such an effect with our technique: The spin polarization measures only the spin part of the magnetization, not the orbital part [129]. We can, however, exclude a drastic reduction of the
magnetic moment for the thicker films, as one would expect for a transition from a high-spin to a low-spin phase, with N 2.2~~ and ,< 1.5~.,, respectively [130,1311. In this respect fee Fe/ Cu(100) behaves analogously to other systems investigated earlier, such as Fe/Au(lll) [84], Fe/ Ag(100) [231 and Co/ A~(1111 [39], which is contrary to an earlier interpretation [23] based on measuring with a macroscopic rather than a microscopic technique. Second, we do not identify a thickness range where both Ms I and Ms,, vanish. During the decay of MS 1 , MS,, already sets in. On the other hand, we see a marked drop of Ms at the switching thickness. I want to point out, however, that this drop occurs within a film thickness interval of only 0.4 ML. By taking high-resolution domain images at this thickness, we find that there are small irregular in-plane domains of several urn size with reversed magnetization exactly at d,. Hence, the apparent reduction in MS might originate from the average over these small domains by large-scale scans and not from a truly reduced magnetization 11321. Here however we approach the limit of the wedge technique: The mutual magnetostatic interaction between adjacent regions of different thickness comes into play, and might influence our observation at d,. A typical stray field range is of the order of some km, which translates into a thickness change of several tenths of a monolayer in our wedge sample. To avoid this possible pitfall, the characteristic
T (K)
230
248
258
266
280
285
295
Fig. 11. Temperature dependence of magnetic domain pattern for an Fe/Cu(lOO) film of constant thickness d,, grown at T = 90 K, and annealed to T = 300 K. Upper panel: magnetization component perpendicular to surface; lower panel: one magnetization component parallel to the surface. At T = 200 K, the film is magnetized by a perpendicular field pulse of H = 5 kA/m to produce a single domain state. Note the change to perpendicular stripe domains during magnetization direction reorientation. Scan area is 50 urn x 50 pm. From [69].
180
drop at d, shown in Fig. 10 must be investigated in films of constant thickness. These results will be discussed in the following. Films of fixed thickness d, have been grown at T = 90 K and annealed to T = 300 K. At low temperatures, these films were made single domain either by applying a short field pulse of 5 kA/m or by growing them in a small field of 250 A/m. Then the evolution of the domain pattern with increasing temperature is directly monitored, see Fig. 11. Below T = 230 K, the film is perpendicularly magnetized and in a single-domain state. At T = 230 K, some irregularly-shaped reversed domains nucleate at a typical distance of N 20 pm, preferably at defects visible in topography. At T = 248 K, the density of nucleation centers has increased, and they tend to be elongated in shape. At T = 258 K, we identify a novel perpendicularly magnetized stripe-domain phase, with stripes running along the [OOl] direction. The stripe distance narrows with increasing temperature. Between 280 K < T G 285 K, the magnetization reorientation transition into plane occurs. It is completed at T = 295 K, where in-plane domains of varying size are visible. We emphasize that the magnetization switching from perpendicular to in-plane occurs over a finite temperature interval of AT, = 50 K. Our data definitely rule out that a loss of long-range ferromagnetic order is responsible for the vanishing of the remanence earlier observed. This qualitative conclusion can be quantified in more detail by an analysis of the images of Fig. 11. For each temperature, the spontaneous magnetization as well as the remanence are determined. The value of M, is approximated by the polarization average over the scan area. Figure 12 demonstrates that M, vanishes at T, and is strongly suppressed in a broad temperature range. MS, on the other hand, does not vanish in such an extended transition and switches in a narrow interval. Note that from Fig. 12 we conjecture that Ms might also be reduced in a much narrower temperature interval of < 10 K at T,, as calculated in Ref. [113]. The investigation of the critical behavior of this remarkable phase transition is a formidable experimental task, which has not yet been tackled. At present we concentrate on the strik-
160
180
200
220
240 (K)
260
280
300
Temperature
Fig. 12. Perpendicular and parallel components of spontaneous magnetization M, and remanent magnetization M, vs. temperature as deduced from the analysis of the images in Fig. 11. M, is taken as the average over the scan area. Note that the drop in M, occurs in a much wider temperature range than the one of Ms.
ing series of domain patterns connected with the reorientation transition in an extended temperature range of 50 K. With rising temperature, a pattern of perpendicular irregular nucleation of domains evolves, followed by perpendicular stripes with narrowing distance, and then by a continuous transition into plane. The most striking observation certainly is the formation of a stripe pattern initiated by the nucleation of small reversed domains, and the preferential orientation of the stripes along the [OOl] direction. Yafet and Gyorgy [81 have predicted that stripe domains are energetically preferred over a uniformly magnetized state owing to the presence of long-range magnetic dipolar interaction. As discussed in Section 4, the stripe period, however, is only accessible to the experiment if surface and shape anisotropies nearly cancel. Otherwise, a single domain should prevail on the length scale of any experiment. The observed transformation of the pattern from single domain at low temperature to a striped configuration near T, qualitatively confirms these predic-
181
tions. This is quite astonishing in view of the fact that the calculation was a zero-temperature approach and that magnetization is assumed to be saturated. The latter hypothesis particularly does not necessarily hold near the phase transition, see Fig. 12. A more detailed approach by Kashuba and Pokrovsky lifts these restrictions by taking the finite-temperature spin fluctuations into account within a renormalization-group analysis [133]. These authors are able to confirm the stripe pattern of Ref. [S]. In addition, they find a preferential orientation of the stripes that is stable against thermal fluctuations. They predict the (001) directions to be favored, in accordance with our observation. The narrowing of the stripes on approaching T, is caused by the strong decrease of the surface anisotropy. The limiting domain width at switching is then given by the dipole length L, a J/E,, which in turn is given by the exchange integral J and the dipolar interaction energy E,. The agreement between the observed minimum width of - 2 km and L, = 1 km [133] is reasonably good, given the fact that a very small external magnetic field is predicted to partly suppress the formation of equally spaced stripes [134]. A remarkable analogy exists between this magnetic stripe phase and a smectic liquid crystal [ 1331. Both phenomena are essentially described by a phase that displays two features, i.e. positional disorder and orientational long-range order. The destruction of positional order can be understood by the displacement of domain walls by thermal fluctuations and dislocation-mediated melting. I would like to briefly discuss a subtlety of the theory concerning the stripe orientation along a crystallographic axis. It is argued that this preferential arrangement along the (001) directions is caused by higher-order terms of the exchange energy E,, which are usually neglected but become relevant in ultrathin films. A similar argument has been advocated by Zhang [135] to explain the thickness dependence of the reorientation temperature. This is quite contrary to the usual description that the crystalline anisotropy EA determines not only the magnetization direc-
tion but also the energetically favored wall orientation. Experimentally we indeed find the in-plane anisotropy direction at T > T, to match the stripe direction at T G T,, which may be more than just a mere coincidence. However, I am not aware of an independent experimental proof, either in favor of the exchange or of the magnetocrystalline anisotropy mechanism. It may very well be that also in this respect an ultrathin film differs considerably from a bulk specimen, and extrapolations from 3D to 2D are generally quite risky. Although so far such a detailed microscopic study has only been undertaken in Fe/Cu(lOO), I expect that other systems displaying a reversible magnetization direction reorientation behave similar. In particular bee-Fe/Ag(lOO) has similar characteristics of a vanishing or at least strongly reduced remanence in a finite temperature range (or likewise in a finite thickness range) [23]. A recent magneto-optic Kerr effect study confirms this main result and has revealed that M, is greatly suppressed but not completely vanishing [24]. Furthermore it identifies an asymmetry within the gap of reduced remanence. However, the study provides no additional insight about the nature of the phase transition, since it also is not spatially resolved. Moreover it suffers from two inherent difficulties of the technique: An unequivocal distinction between different magnetization components is very hard to achieve, and small misalignments of the external field govern the shape of the hysteresis loop [24]. Further experiments are necessary to test the theoretical predictions of this reorientation phase transition in detail. In particular the observed asymmetry between oppositely oriented stripe domains is not completely understood [69,133]. Moreover it has been argued [llO] that the switching temperature T, is not uniform. Within the domain walls, vertical magnetization components might be preserved up to temperatures above T,, a prediction that stimulates a careful investigation on the domain walls during this phase transition. Finally I expect that such a magnetization reorientation with a characteristic domain configuration occurs not only with film thickness and temperature, but can also be induced by an applied magnetic field [113]. For a
182
technique based on polarized electrons, such as spin-SEM, domain observation in external fields is not a trivial task. For a study in an external field, on the other hand, magnon light scattering [136] is the paradigm experiment, because of its ability to tune the scattering vector of the light to the spatial frequency of interest - in this case the stripe-domain period.
7. Conclusions This work was not intended as a systematic overview of all studies in the exploding field of thin film magnetism. My intention was to illustrate some of the phenomena encountered in ultrathin films for which a spatially resolving technique can lead to a deeper understanding. First I began by discussing some very basic concepts on domain formation in general, and specialized this approach to ultrathin films. The first topic was a comprehensive discussion of why and when magnetic domains form in these films and how the experimental observations relate to existing theoretical work. The role of thermal fluctuations below and above the Curie temperature was investigated and discussed in view of the strikingly different properties of a ferromagnet in 2D as compared to 3D. In particular, owing to the high spatial resolution of spin-SEM, we were able to distinguish between microscopic and macroscopic quantities, the spontaneous and the remanent magnetization. This unique feature of the technique has also been exploited for the an investigation on how the third subject, anisotropies influence ferromagnetic long-range order. The properties of ultrathin films can be tailored in such a way that the competition between surface and shape anisotropies almost leads to a compensation of these terms. We have investigated such a system in detail, and approached that realm of 2D magnetism in which anisotropies are largely suppressed. Hitherto unobserved types of magnetic domain patterns have been identified and attributed to the 2D nature of these films. I am aware that some fascinating topics in 2D magnetism and especially in micromagnetism have been omitted in this work. In particular, the high
resolution capability of spin-SEM which allows magnetization profiles to be measured across domain walls, has not been included [46,137,138]. Moreover we have restricted ourselves to single ultrathin films, with the intention to match the experimentalists dream and the theorists reality of the free-standing magnetic monolayer as closely as possible. Therefore I did not discuss exchange-coupled layers and the marvellous aesthetic contributions spin-SEM provides in visualizing the oscillations of this exchange coupling with interlayer spacing [122]. All these different physical aspects of 2D magnetism on the microscopic scale make up the fascination of this field of physics: Through hard experimental work, some answers can be found, which in turn pose even more new questions. I am convinced that 2D magnetism still has more secrets to be unveiled.
Acknowledgment Since the start of this domain-imaging project several years ago, I have profited from the invaluable help of many persons, too numerous to list them all here. Representative of them is my colleague, Andi Bischof, to whom I take pleasure in giving special thanks for his major contributions in designing, constructing, repairing and perfecting the equipment. The content of this paper has been presented as the 1992 Wohlfarth lecture at the CMMP 92 Conference in Sheffield. I would like to thank the IOP Magnetism Group for electing me to give this lecture and Elsevier for sponsoring that event.
References
[ll P. Weiss, J. de Physique 6 (1907) 661. [2] L. Landau and E. Lifshitz, Physik. Z. Sowjetunion 8 (1935) 153. [3] D.J. Craik and R.S. Tebble, in: Ferromagnetism and Ferromagnetic Domains, ed. E.P. Wohlfarth (NorthHolland, Amsterdam, 1965). [4] R. Carey and E.D. Isaac, Magnetic Domains and Techniques for their Observation (Academic Press, New York, 1966).
R. AIlenspach /Journal
of Magneiism and Magnetic Materials 129 (1994) 160-185
183
51 C. Kittel, Phys. Rev. 70 (1946) 965. 51 2. M&k and V. Kamberslj, Czech. J. Phys. 8 (1958) 416. [7] C. Kooy and U. Enz, Philips Res. Rep. 15 (1960) 7. [8] Y. Yafet and E.M. Gyorgy, Phys. Rev. B 38 (1988) 9145. [9] R. Czech and J. Villain, J. Phys. Condens. Matter 1 (1989) 619. [lo] K. Koike, H. Matsuyama and K. Hayakawa, Scanning Microsc. Suppl. 1 (1987) 241. [ll] M.R. Scheinfein, J. Unguris, M.H. Keliey, D.T. Pierce and R.J. Celotta, Rev. Sci. Instrum. 61 (1990) 2501. [12] T.H. DiStefano, IBM Techn. Discl. Bull. 20,4212 (March 1978). [13] H.C. Siegmann, J. Phys. Condens. Matter 4 (1992) 8395. 1141 M. Taborelli, Ph.D. Thesis No. 8545, ETH, Zurich, Switzerland (1988). [15] D.P. Pappas, K.-P. Kamper, B.P. Miller, H. Hopster, D.E. Fowler, CR. Brundle, A.C. Luntz and Z.-X. Shen, Phys. Rev. Lett. 66 (1991) 504. [16] M. Landolt, unpublished. [17] R. Allenspach, A. Bischof, J. Mannhart and K. Wasser, unpublished. [18] S.D. Bader, J. Magn. Magn. Mater. 100 (1991) 440. [19] D. Kerkmann, Appl. Phys. A 49 (1989) 523. [20] A. Bischof, Diploma Thesis, Juventus, Zurich, Switzerland (19931. [21] J.L. Robins, R.J. Celotta, J. Unguris, D.T. Pierce, B.T. Jonker and G.A. Prinz, Appl. Phys. Lett. 52 (1988) 1918. [22] M. Stampanoni, A. Vaterlaus, M. Aeschlimann and F. Meier, Phys. Rev. Lett. 59 (1987) 2483. [23] D.P. Pappas, C.R. Brundle and H. Hopster, Phys. Rev. B 45 (1992) 8169. [24] Z.Q. Qiu, J. Pearson and S.D. Bader, Phys. Rev. Lett. 70 (1993) 1006. [25] H.P. Oepen, M. Benning, H. Ibach, C.M. Schneider and J. Kirschner, J. Magn. Magn. Mater. 86 (1990) L137. [26] C. Chappert, D. Renard, P. Beauvillain, J.P. Renard and J. Seiden, J. Magn. Magn. Mater. 54-57 (1986) 795. 1271 F.J.A. den Broeder, D. Kuiper, A.P. van de Mosselaer and W. Hoving, Phys. Rev. Lett. 60 (1988) 2769. ]28] C.H. Lee, H. He, F. Lamelas, W. Vavra, C. Uher and R. Clarke, Phys. Rev. Lett. 62 (1989) 653. [29] C.H. Lee, H. He, F.J. Lamelas, W. Vavra, C. Uher and R. Clarke, Phys. Rev. B 42 (1990) 1066. [30] P.F. Garcia, A.D. Meinhaldt and A. Suna, Appl. Phys. Lett. 47 (1985) 178. [31] P.F. Garcia, J. Appl. Phys. 63 (1988) 5066. 1321 W.B. Zeper, F.J.A.M. Greidanus, P.F. Garcia and C.R. Fincher, J. Appl. Phys. 65 (1989) 4971. [33] C.H. Lee, R.F.C. Farrow, C.J. Lin, E.E. Mariner0 and C.J. Chien, Phys. Rev. B 42 (1990) 11384. 1341 L. Neel, J. Phys. Rad. 15 (1954) 376. [35] C.J. Chien, B.M. Clemens, S.B. Hagstrom, R.F.C. Farrow, C.H. Lee, E.E. Mariner0 and C.J. Lin, Mat. Res. Sot. Symp. Proc. 231 (1992) 465. 1361 R.H. Victora and J.M. MacLaren, Phys. Rev. B. 47 (19931 11583.
[37] F.J.A. den Broeder, W. Hoving and P.J.H. Bloemen, J. Magn. Magn. Mater. 93 (1991) 562. [38] R. Jungblut, M.T. Johnson, J. aan de Stegge, A. Reinders and F.J.A. den Broeder, unpublished. [39] R. Allenspach, M. Stampanoni and A. Bischof, Phys. Rev. Lett. 65 (1990) 3344. [40] G. Bayreuther, P. Bruno, G. Lugert and C. Turtur, Phys. Rev. B 40 (1989) 7399. [41] J. FerrC, G. Penissard, C. Mar&e, D. Renard, P. Beauvillain and J.P. Renard, Appl. Phys. Lett. 56 (19901 1588. [42] Taking into account the Co bulk parameters and the film thickness of 3 ML in the calculation of Ref. [B] leads to slightly modified values as compared to the original work. [43] J.N. Chapman, S. McVitie and I.R. McFadyen, Scanning Microsc. Suppl. 1 (1987) 221. [44] R. Ploessl, J.N. Chapman, M.R. Scheinfein, J.L. Blue, M. Mansuripur and H. Hoffmann, unpublished. [45] C. Chappert and P. Bruno, J. Appl. Phys. 64 (1988) 5736. [46] R. Allenspach and M. Stampanoni, Mat. Res. Sot. Symp. Proc. 231 (1992) 17. [47] B. Voigtllnder, G. Meyer and N.M. Amer, Phys. Rev. B 44 (1991) 10354. [48] Ch. W611, S. Chiang, R.J. Wilson and P.H. Lippel, Phys. Rev. B 39 (1989) 7899. [49] J. Pommier, P. Meyer, G. Penissard, J. FerrC, P. Bruno and D. Renard, Phys. Rev. Lett. 65 (1990) 2054. [50] The earths magnetic field is not shielded; J. Ferre, private communication. 1511 P. Griitter and U. Diirig, unpublished. [52] D. Pescia, M. Stampanoni, G.L. Bona, A. Vaterlaus, R.F. Willis and F. Meier, Phys. Rev. Lett. 58 (1987) 2126. [53] D. Pescia, M. Stampanoni, G.L. Bona, A. Vaterlaus, F. Meier, G. Jennings and R.F. Willis, Phys. Rev. Lett. 60 (1988) 2559. [54] D.A. Steigenvald, I. Jacob and W.F. Egelhoff, Jr., Surf. Sci. 202 (1988) 472. 1551 D.A. Steigerwald and W.F. Egelhoff, Jr., Phys. Rev. Lett. 60 (1988) 2558. ]56] H. Magnan, D. Chandesris, B. Villette, 0. Heckmann and J. Lecante, Phys. Rev. Lett. 67 (1991) 859. [57] M. Wuttig and J. Thomassen, Surf. Sci. 282 (1993) 237. 1581 S.H. Lu, J. Quinn, D. Tian, F. Jona and P.M. Marcus, Surf. Sci. 209 (1989) 364. [59] M. Stampanoni and R. Allenspach, J. Magn. Magn. Mater. 104-107 (1992) 1805. 1601 D.P. Pappas, K.-P. Kiimper, H. Hopster, D.E. Fowler, A.C. Luntz, C.R. Brundle and Z.-X. Shen, J. Appl. Phys. 69 (1991) 5209. [611 J. Thomassen, B. Feldmann and M. Wuttig, Surf. Sci. 264 (19921406. [62] P. Xhonneux and E. Courtens, Phys. Rev. B 46 (1992) 556.
84
of Magnetism and Magnetic Materials 129 (1994) 160-185 [90] Z.Q. Qiu, S.H. Mayer, C.J. Gutierrez, H. Tang and J.C. Walker, Phys. Rev. Lett. 63 (1989) 1649. [91] A.M. Begley, D. Tian, F. Jona and P.M. Marcus, Surf. Sci. 280 (1993) 289. [92] B. Voigtlinder, G. Meyer and N.M. Amer, Surf. Sci. 255 (1991) L529. [93] J.A. Stroscio, D.T. Pierce, R.A. Dragoset and P.N. First, J. Vat. Sci. Technol. A 10 (1992) 1981. [94] V.L. Pokrovsky, Advances in Physics 28 (1979) 595. [95] P. Weiss and R. Forrer, Compt. Rend. 178 (1924) 1670. [96] S.B. Khokhlachev, Sov. Phys. JETP 44 (1976) 427. [97] D.J. Webb and J.D. McKinley, Phys. Rev. Lett. 70 (19931509. [98] N.C. Koon, G.A. Prinz, F.A. Volkening, B.N. Das and K. Aylesworth, J. Appl. Phys. 69 (1991) 5967. [99] C.H. Back, A. Kashuba and D. Pescia, unpublished. [loo] G.A. Mulhollan, R.L. Fink, J.L. Erskine and G.K. Walters, Phys. Rev. B 43 (1991) 13645. [loll Yu.S. Karimov, Sov. Phys. JETP 38 (1974) 129. [102] K. Hirakawa and K. Ubukoshi, J. Phys. Sot. Jpn. 50 (1981) 1909. [103] S.O. Demokritov, N.M. Kreines, V.I. Kudinov and S.V. Petrov, Sov. Phys. JETP 68 (1989) 1277. [104] For a thorough discussion of the procedures and fits to extract critical exponents from experimental results and of the possible pitfalls, see Ref. [79]. [105] H. Godfrin, R.R. Rue1 and D.D. Osheroff, Phys. Rev. Lett. 60 (1988) 305. [106] L.J. Friedman, A.L. Thomson, C.M. Gould, H.M. Bozler, P.B. Weichman and M.C. Cross, Phys. Rev. Lett. 62 (1989) 1635. [107] H. Godfrin, R.R. Rue1 and D.D. Osheroff, Phys. Rev. Lett. 69 (1992) 1288. [108] D. Kerkmann, D. Pescia and R. Allenspach, Phys. Rev. Lett. 69 (1992) 1289. [109] N.D. Mermin and H. Wagner, Phys. Rev. Lett. 17 (1966) 1133. [llO] P.J. Jensen and K.H. Bennemann, Phys. Rev. B 42 (19901 849. [ill] D. Pescia and V.L. Pokrovsky, Phys. Rev. Lett. 65 (1990) 2599. [112] P.J. Jensen and K.H. Bennemann, Solid State Commun. 83 (19921 1057. [113] R.P. Erickson and D.L. Mills, Phys. Rev. B 46 (1992) 861. [114] J. Araya-Pochet, C.A. Ballentine and J.L. Erskine, Phys. Rev. B 38 (1988) 7846. [115] D.P. Pappas, K.-P. Kamper and H. Hopster, Phys. Rev. Lett. 64 (1990) 3179. [116] M.T. Kief, G.J. Mankey and R.F. Willis, J. Appl. Phys. 70 (1991) 5929. [117] N.C. Koon, B.T. Jonker, F.A. Volkening, J.J. Krebs and G.A. Prinz, Phys. Rev. Lett. 59 (19871 2463. [118] F.A. Volkening, B.T. Jonker, J.J. Krebs, N.C. Koon and G.A. Prinz, J. Appl. Phys. 63 (19881 3869. [119] Note the difference by a factor of 2 in the KS term due
[63] D.D. Chambliss, R.J. Wilson and S. Chiang, J. Vat. Sci. Technol. A 10 (1992) 1993. 1641 A. Brodde and H. Neddermeyer, Surf. Sci. 287/288 (1993) 988. [65] Th. Detzel, N. Memmel and Th. Fauster, Surf. Sci 293 (1993) 227. [66] B. Kaplan and G.A. Gehring, J. Magn. Magn. Mater. 128 (19931 111. [67] Z.-P. Shi, .I. Phys. Condens. Matter 4 (1992) L191. 1681 R. Allenspach, A. Bischof, M. Stampanoni, D. Kerkmann and D. Pescia, Appl. Phys. Lett. 60 (1992) 1908. [69] R. Allenspach and A. Bischof, Phys. Rev. Lett. 69 (1992) 3385. 1701 See, e.g., S. Chikazumi, Physics of Magnetism (Wiley, New York, 1964). 1711 A very illuminating discussion on 3D ferromagnetism at the Curie temperature can be found in W. Gerlach, Sci. Rep. Tohoku Imperial University (I) (1936) 248. 1721 D. Mauri, D. Scholl, H.C. Siegmann and E. Kay, Phys. Rev. L&t. 62 (1989) 1900. ]73] D. Mauri, D. Scholl, H.C. Siegmann and E. Kay, Appl. Phys. A 49 (1989) 439. [74] ST. Bramwell and P.C.W. Holdsworth, J. Phys. Condens. Matter 5 (1993) L53. [75] W. Diirr, M. Taborelli, 0. Paul, R. Germar, W. Gudat, D. Pescia and M. Landolt, Phys. Rev. Lett. 62 (1989) 206. [76] C. Liu and S.D. Bader, J. Appl. Phys. 67 (1990) 5758. [77] M. Farle and K. Baberschke, Phys. Rev. Lett. 58 (1987) 511; K. Baberschke, M. Farle and M. Zomack, Appl. Phys. A 44 (1987) 13. [78] C. Rau, Appl. Phys. A 49 (1989) 579. [79] J. Kohlhepp, H.J. Elmers, S. Cordes and U. Gradmann, Phys. Rev. B 45 (1992) 12287. [80] D. Kerkmann, D. Pescia and R. Allenspach, Phys. Rev. Lett. 68 (1992) 686. [81] A.K. Schmid and J. Kirschner, Ultramicroscopy 42-44 (1992) 483. [82] H.E. Stanley, Introduction to Phase Transitions and Critical Phenomena (Clarendon Press, Oxford, 1971). [83] D. Pescia, G. Zampieri, M. Stampanoni, G.L. Bona, R.F. Willis and F. Meier, Phys. Rev. Lett. 58 (1987) 933. [84] G. Lugert, G. Bayreuther, S. Lehner, G. Gruber and P. Bruno, Mat. Res. Sot. Symp. Proc. 232 (1992) 97. 1851 M. Farle, K. Baberschke, U. Stetter, A. Aspelmeier and F. Gerhardter, Phys. Rev. B 47 (1993) 11571. [86] D. Stauffer, AIP Conf. Proc. 10 (1972) 827. [87] U. Gradmann, M. Przybylski, H.J. Elmers and G. Liu, Appl. Phys. A 49 (1989) 563. [88] For an introduction to superparamagnetism, see e.g. the textbook by B.D. Cullity, Introduction to Magnetic Materials, Addison-Wesley, Reading, 1972; for the original article, C.P. Bean, J. Appl. Phys. 26 (19551 1381. [89] I.S. Jacobs and C.P. Bean, in: Magnetism, Vol. 3, p. 271, eds. G.T. Rado and H. Suhl (Academic Press, New York, 1963).
185
[120] [121] [122] [123] [124] [125] [126] [127]
[128] [129]
to the fact that Ref. [45] deals with Au/Co/Au(lll) sandwiches while we use a Co/Au(lllI film and hence have half the number of interfaces. S. Methfessel, S. Middelhoek and H. Thomas, J. Appl. Phys. 31 (1960) 302s. P. Griinberg, S. Demokritov, A. Fuss, M. Vohl and J.A. Wolf, J. Appl. Phys. 69 (1991) 4789. J. Unguris, R.J. Celotta and D.T. Pierce, Phys. Rev. Lett. 67 (1991) 140. J. Thomassen, F. May, B. Feldmann, M. Wuttig and H. Ibach, Phys. Rev. Lett. 69 (1992) 3831. W.A.A. Macedo, W. Keune and E.D. Ellerbrock, J. Magn. Magn. Mater. 93 (1991) 552. C. Liu, E.R. Moog and S.D. Bader, Phys. Rev. Lett. 60 (1988) 2422. M. Stampanoni, Appl. Phys. A 49 (1989) 449. R. Allenspach, in: Magnetism and Structure in Systems of Reduced Dimension, eds. R.F.C. Farrow et al., NATO AS1 Series B 309 (Plenum, New York-London, 19931, p. 33. P. Bruno, Ph.D. Thesis, Universite de Paris-Sud, Orsay, France (1989). The orbital component of the magnetization does not contribute to the spin polarization because averaging occurs during electron emission, which leads to a disappearance of this part. For more details, see Ref. [13].
[I301 C.S. Wang, B.M. Klein and H. Krakauer, Phys. Rev. Lett. 54 (1985) 1852. [131] V.L. Moruzzi, P.M. Marcus, K. Schwarz and P. Mohn, Phys. Rev. B 34 (1986) 1784. [132] In other words, our concept of measuring the microscopic MS breaks down: When averaging across domain boundaries, we instead measure a remanent magnetization, averaged over the pixel distance or beam diameter, whichever is larger. [133] A. Kashuba and V.L. Pokrovsky, Phys. Rev. Lett. 70 (1993) 3155; Phys. Rev. B 48 (1993) 10335. [134] In this study the magnetic field during spin-SEM operation is = 100 A/m, owing to the objective lens pole-piece of the electron microscope. This small field might already be sufficient to produce the asymmetry observed between the widths of black and white stripe domains in Fig. 11, as discussed in Ref. [133]. [135] S. Zhang, Phys. Rev. B 45 (1992) 7487. [136] For a review, see e.g. P. Griinberg, Progress in Surface Science 18 (1985) 1. [137] A. Berger and H.P. Oepen, Phys. Rev. B 45 (1992) 12596. [138] A. Berger, Ph.D. Thesis, RWTH Aachen, Germany (1993).

Allenspach

Uploaded by

Copyright:

Available Formats

Allenspach

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Allenspach

Uploaded by

Copyright:

Available Formats

JTsgL&

Journal of Magnetism and Magnetic Materials 129 (1994) 160-185

Ultrathin films: magnetism on the microscopic scale

(Received 22 July 1993)

0 1994 Elsevier Science B.V. All rights reserved

of Magnetism and Magnetic Materials 129 (1994) 160-185

2. Why do magnetic domains

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-18.5

4. Experimental thin films

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

image of a 4-ML hcp-Co/Pt(lll) domains with magnetization

of Magnetism and Magnetic Materials 129 (1994) 160-185

R. Allenspach/Journal of Magnetismand MagneticMaterials129 (1994) 160-185

of Magnerism and Magnetic Materials 129 (1994) 160-185

5. Two-dimensional magnetism the Curie temperature

below and above

of Magnetism and Magnetic Materials I29 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

ing on the actual 6 = o, 0 = 90, and sin20 = (~/2

3.0 4.0 Film Thickness

6.0 5.0 (ML)

of Magnetism and Magnetic Materials 129

ofMagnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

of Magneiism and Magnetic Materials 129 (1994) 160-185

of Magnetism and Magnetic Materials 129 (1994) 160-185

[120] [121] [122] [123] [124] [125] [126] [127]

You might also like