MATERIALS REPORTS Surface, interface, and thin-film magnetismischuller.ucsd.edu/pdfs/Paper...

MATERIALS REPORTSTechnical reports giving an overview of progress and challenges in areas of materials research will be included in thissection periodically.

Surface, interface, and thin-film magnetismL. M. FalicovDepartment of Physics, University of California-Berkeley, and Lawrence Berkeley Laboratory,Berkeley, California 94720

Daniel T. PierceNational Institute of Standards and Technology, Gaithersburg, Maryland 20899

S. D. BaderMaterials Science Division, Argonne National Laboratory, Argonne, Illinois 60439

R. GronskyDepartment of Materials Science and Mineral Engineering, University of California-Berkeley, andLawrence Berkeley Laboratory, Berkeley, California 94720

Kristl B. HathawayOffice of Naval Research, Arlington, Virginia 22217-5000

Herbert J. HopsterDepartment of Physics, University of California-Irvine, Irvine, California 92717

David N. LambethDepartment of Electrical and Computer Engineering, Carnegie Mellon University, Pittsburgh,Pennsylvania 15213-3890

S. S. P. ParkinIBM. Almaden Research Center, San Jose, California 95120-6099

Gary PrinzNaval Research Laboratory, Washington, DC 20375

Myron SalamonDepartment of Physics, University of Illinois, Urbana, Illinois 61801

Ivan K. SchullerDepartment of Physics, University of California-San Diego, La Jolla, California 92093

R. H. VictoraDiversified Technologies Research Laboratories, Eastman Kodak Company, Rochester,New York 14650-2017

(Received 19 October 1989; accepted 7 March 1990)

A comprehensive review and state of the art in the field of surface, interface, andthin-film magnetism is presented. New growth techniques which produce atomicallyengineered novel materials, special characterization techniques to measure magneticproperties of low-dimensional systems, and computational advances which allow largecomplex calculations have together stimulated the current activity in this field andopened new opportunities for research. The current status and issues in the area ofmaterial growth techniques and physical properties, characterization methods, andtheoretical methods and ideas are reviewed. A fundamental understanding of surface,interface, and thin-film magnetism is of importance to many applications in magneticstechnology, which is also surveyed. Questions of fundamental and technological interestthat offer opportunities for exciting future research are identified.

J. Mater. Res., Vol. 5, No. 6, Jun 1990 1299

L. M. Falicov, D.T. Pierce et al.\ Surface, interface, and thin-film magnetism

TABLE OF CONTENTS

I. INTRODUCTION 1301

II. THEORETICAL BACKGROUND 1301A. Electronic structure 1301B. Phenomenology and model systems . . . 1302C. Critical phenomena theories 1303D. Transport properties in

magnetic systems 1303E. Micromagnetic theory 1303

III. MATERIALS 1304A. Growth techniques 1304

1. Sputtering 13042. Molecular beam epitaxy (MBE).... 13043. Metal-organic chemical

vapor deposition (MOCVD) 13054. Production techniques 1305

B. Growth modes 1305C. Systems highlights 1306

1. Surface and monolayer films 13072. Metastable epitaxial films 13083. Semiconductor substrates 13094. Rare earths 13095. Oxides 13106. Multilayers 1310

D. Physical properties 13101. Proximity and interfacial effects . . . 13102. Exchange coupling across

interfaces 13113. Coupling through nonmagnetic

layers 13124. Ruderman-Kittel-Kasuya-Yosida

(RKKY) coupling 13135. Magnetoelasticity 13136. Superlattice effects 1314

IV. TECHNIQUES AND FACILITIES 1315A. Magnetometry-spectroscopy 1315

1. Mossbauer spectroscopy 13152. Magnetic resonance 13163. Magnetometry 13164. Magneto-optics 1317

B. Polarized electron techniques 13171. Spin-polarized photoemission

spectroscopy 13172. Polarized Auger spectroscopy 1318

3. Spin-polarized low-energyelectron diffraction (SPLEED) 1319

4. Spin-polarized secondaryelectron emission (SPSEE) 1319

5. Spin-polarized electron energy-lossspectroscopy (SPEELS) 1320

6. Polarized particle probes 1320C. Electron microscopy 1320

1. Scanning electron microscopy withpolarization analysis (SEMPA) 1321

2. Lorentz electron microscopy 13213. Conventional electron microscopy.. 13224. Scanning tunneling microscopy

(STM) and magnetic forcemicroscopy (MFM) 1323

D. Diffraction 13231. X-ray diffraction 13232. Neutron scattering 1324

E. Photon sources 1325

V. APPLICATIONS 1326A. Magnetic recording 1326

1. Hard disk media 13272. Magneto-optic media 13273. Exchange biasing of

magnetoresistive heads 13274. High magnetization materials

for recording heads 1328B. Magnetoelastic devices 1328C. Integrated optical and

electronic devices 1328D. Permanent magnets 1329

VI. ISSUES AND PROSPECTS 1329A. Theory 1329B. Magnetic moments at

surfaces and interfaces 1330C. Magnetic coupling at interfaces 1330D. Low-dimensional magnetism 1331E. Excitations 1331F. Magnetism and structure 1332G. Metastability 1333H. Anisotropy and magnetostriction 1333I. Magnetoresistivity 1333J. Micromagnetics 1334K. Magnetics technology 1335L. Conclusion 1335

VII. REFERENCES 1336

1300 J. Mater. Res., Vol. 5, No. 6, Jun 1990

L.M. Falicov, D.T. Pierce etai: Surface, interface, and thin-film magnetism

I. INTRODUCTION

With the information revolution and the ever-growing need to acquire, store, and retrieve informa-tion, the science and technologies attached to magneticrecording have experienced an explosive growth. Centralto those pursuits is the materials science of magnetismas it applies to surfaces, interfaces, and thin films.

Magnetism is an electronically driven phenomenon,weak compared with electrostatic effects but subtle inits many manifestations. It is quantum-mechanical innature, with its origins in the Pauli exclusion principleand the existence of the electron spin. It leads, nonethe-less, to a large variety of short- and long-range forces,and both classical and quantum-mechanical effects.This last feature provides the richness of textures andproperties encountered in magnetic systems, from whichuseful engineering and technical applications arise.

The preparation and especially the control of sur-faces and interfaces in magnetic systems open a newarea in the science of magnetism, one that involves ahighly interdisciplinary endeavor: physics, chemistry,and materials sciences; theory and experiment; surfacescience; small laboratory and central-facility research;materials preparation and characterization; academic,national-laboratory, and industrial research.

The present report is the result of the deliberationsof a Panel convened in Santa Fe, New Mexico, on June18-21, 1989, under the auspices of the Department ofEnergy, Council on Materials Science. The Panel, oftwelve members, was chaired by Falicov and Pierce.The Panel's charge was to assess the state of the art inthe area, identify the major, important issues, and esti-mate the prospects for future research.

Several technical developments are responsible forthe intense activity in the field. In addition to theapplication-driven pressures mentioned above, threemajor advances are to be noted:

(1) The advent of new sample-preparation techniqueswhich now permit the manufacture of single-purposedevices to extraordinarily accurate specifications; thesetechniques [Molecular Beam Epitaxy (MBE), Metal-Organic Chemical Vapor Deposition (MOCVD), sput-tering, lithography, etc.] are becoming increasinglyavailable and less expensive and have engendered, inaddition to the obvious technological progress, a newbranch of "pure" science concerned with artificiallymade systems.

(2) The availability of better and sophisticatedsample characterization techniques, based mostly (al-though not exclusively) on centrally located facilities.These techniques are based on x-ray and ultra-violetphotons (synchrotron sources), visible and infrared pho-tons (ordinary and free-electron lasers), neutrons (reac-tors and pulsed neutron sources), and electrons of avariety of energies (electron microscopes of several

kinds; low-, intermediate-, and high-energy electronsources for elastic and inelastic scattering experiments).To these should be added the existence and readyavailability of excellent controlled environments (goodvacuum and clean gaseous atmospheres; from very lowto very high temperatures; high and spatially uniformmagnetic fields).

(3) The increasing availability of fast, operationallyinexpensive and numerically intensive computers whichhave permitted the calculation of a large variety ofproblems related to realistic systems, in complicatedgeometries, with subtle quantum-mechanical effects,and/or for practical devices.

This combination of factors makes it almost a neces-sity to evaluate, even though briefly, where the fieldis, where it is going, where the needs are greater, andwhere the better pay-offs may lie.

II. THEORETICAL BACKGROUND

Theoretical techniques relevant to understand-ing magnetic phenomena at surfaces, interfaces, andin thin films are grouped into five general areas:electronic-structure techniques, phenomenological andmodel system theories, theories of critical phenomena,transport theory, and the special phenomenology ofmicromagnetics.

A. Electronic structure

Electronic-structure techniques compute the groundstate of a many-electron solid at zero temperature. Ahierarchy of techniques exists in which successivelymore terms in the Hamiltonian are approximated fromphysical models or experimental data. Current ab initiotechniques require only the specification of atomic posi-tions and species to determine the ground-state energy.They typically use the local density approximation tothe density functional formalism. With state-of-the-artsupercomputers, calculations can be performed for upto ten-atom ordered unit cells and for free surfaces andinterfaces. The following magnetic properties havebeen or can be calculated:

(1) Because magnetic energies are much smallerthan binding energies, it is necessary to solve the struc-tural problem from the outset. Physical structures maybe optimized by comparing the total energies for a se-ries of atomic configurations. Bulk lattice parametersare predicted1 generally to within 1%, and elastic con-stants to within 10%. It should be noted, however, thatthe bulk lattice constants predicted for the magnetic 3dtransition metals2 are too small by as much as 3%. Phys-ical structure determinations for thin films, surfaces,and interfaces are straightforward but extremely time-consuming because of the reduced symmetry. Early cal-culations for thin films and surfaces did not allow for


L.M. Falicov, D.T. Pierce et a/.: Surface, interface, and thin-film magnetism

interlayer relaxation; recent calculations, as discussedbelow, indicate that such relaxations can significantlyaffect the computed magnetic properties. More compli-cated surface reconstructions remain to be explored.

(2) Magnetic moments of bulk transition metalsand some ordered alloys are typically calculated3 towithin 2%. Rare-earth ions can be treated by thesetechniques only if the /-shell configurations are prop-erly constrained.4 Calculations of moments at surfaces,interfaces, and in few-monolayer films, if the positionsof the atoms are correctly specified, can be expectedto have the same accuracy as the bulk moments. In par-ticular, the magnetic moments of surfaces and few-monolayer films have been predicted to be significantlyenhanced5 and, for some normally nonmagnetic materi-als, surface layers are predicted to acquire a magneticmoment.6 However, these calculations have of necessityassumed atomic spacings close to the bulk values;i.e., structural relaxations have not been included. Acalculation for Fe/W, allowing surface layer spacings torelax to their minimum energy configuration, shows anearly total disappearance of the enhanced magneticmoment.7

(3) Magnetic structures can be predicted by compar-ing total energies for a limited set of magnetic structures(which exclude any spin canting) calculable by thesetechniques. For example, the possible antiferromag-netic phases of bulk manganese have been calculated.8In general such calculations agree with experimentalresults, with the notable exception that an antiferro-magnetic face-centered cubic phase is erroneously pre-dicted by the most accurate calculational techniques asthe stable phase of iron.9

(4) Calculated Fermi surfaces of magnetic metalsshow good agreement with experiment10 in some cases(Fe), not as good11"13 for others (Co, Ni). ReliableFermi surfaces are necessary for predicting transportproperties.

Systematic studies of a wide variety of physical andmagnetic structures of surfaces and interfaces currentlyrequire more approximate methods of electronic struc-ture calculations. If such methods are constructed toreproduce known experimental or ab initio results, pre-dictions can be expected to be quite reliable.14

Ab initio methods based on the local density ap-proximation replace real electron correlation potentialsand energies by average values from a homogenouselectron gas, thus effectively giving a one-electron de-scription. For certain systems in which electron correla-tions are important in determining the magnetism, anunderstanding of the electronic structure may be ob-tained only through explicit many-body techniques.

New techniques which combine molecular dynam-ics simulations with ab initio electronic structure deter-minations are still in their infancy,15 but may be

expected, in the long term, to be applied to realisticmagnetic systems to determine the physical and mag-netic structures simultaneously.

B. Phenomenology and model systems

There are several properties of magnetic materialswhich, although derived from the electronic structure,are not adequately treated by current electronic-structuretechniques; these frequently omit the relativistic spin-orbit coupling terms. (Spin-orbit coupling terms havebeen included in ab initio electronic structure calcula-tions, but usually only for closed shell, i.e., nonmag-netic systems.) Spin-orbit related energies are usuallyseveral orders of magnitude smaller than those associ-ated with changes in physical or magnetic structure.These properties, which include anisotropy, magneto-striction, and magneto-optic coupling, have historicallybeen treated by phenomenological models in which theform of the required terms in the Hamiltonian is con-strained by symmetry and the magnitude of the rele-vant coefficients is extracted from physical models andexperimental results.

Magnetic anisotropy is the energy associated with aspecific orientation of the magnetic moment relative tothe crystal axes or macroscopic structure. Work on itin-erant electron systems, where the anisotropy derivesfrom the entire Fermi surface, relies on empirical mod-els. (Early calculations deriving anisotropy from tight-binding band structures, including the spin-orbit termas a perturbation, were in only fair agreement withexperiment.16) For rare-earths and transition-metaloxides, where the anisotropy is associated with localmoments, crystal-field methods have been successful;they, however, almost invariably include some adjust-able parameters.17 Empirical models for anisotropies atsurfaces were developed long ago,18 but even today theyrequire experimental parameters.

Magnetostriction is a change in shape of a body onthe application of a magnetic field. Linear magne-tostriction is the coupling between the direction of themoment and elastic strain; i.e., it is the strain derivativeof the anisotropy energy. Models for magnetostrictionat surfaces are intimately related to surface anisotropy,but have not received much attention. The mismatchstrain at an interface has a large effect on magneticproperties through magnetostriction.

Magneto-optical effects arise from the coupling be-tween the spin and charge polarization, again a spin-orbit effect. Magneto-optical coefficients have usuallybeen derived from experimental results.

Extension of phenomenological theories of spin-orbit related properties to surfaces and interfaces usu-ally requires the inclusion of lower symmetry terms inthe model Hamiltonians. Temperature dependence can


L.M. Falicov, D.T. Pierce etal.: Surface, interface, and thin-film magnetism

be included in these theories by making, phenomeno-logically, the coefficients depend on temperature.Model theories have been developed which relate thetemperature dependences of anisotropy and magneto-striction to the temperature dependence of the mo-ment; they are reasonably successful for local-momentsystems.

C. Critical phenomena theories

Surface, interface, and thin-film magnetism pro-vide a fertile ground to explore critical phenomena,in particular those that arise in response to restricteddimensionality, finite-size effects, and surface-drivenmechanisms. It is well known that, in two dimensions,systems with one degree of freedom (i.e., Ising-like)have a well-defined phase transition.19 In fact, thethermodynamics of that problem, solved exactly byOnsager more than 35 years ago, is one of the landmarksin the Theory of Phase Transitions. It predicts, for thesimple square lattice, a transition temperature Tc equal,in energy units, to 2.269 times the value of the nearest-neighbor Ising exchange parameter. It also finds19"22

that, as the temperature T approaches Tc from below,the magnetization of the system decays to zero as

M = M0[l - (T/Tc)]0125.

In contrast, isotropic systems with two or three degreesof freedom (i.e., xy- and Heisenberg-like) exhibit nolong-range order in two dimensions21"23 at any finitetemperature T. The development of three-dimensionalorder as such systems are built up layer by layer has alsobeen studied.24

There are, in addition, fascinating surface effectsrelated to a variety of critical phenomena: behavior andtransitions involving the decay in short-range order23

(the so-called Kosterlitz-Thouless transition), the inter-play between surface and bulk effects25 (including thepersistence of order on the surface at temperatureshigher than the bulk Curie or Neel temperatures andvarious temperature dependences of the magnetizationof the surface layers as compared to the bulk), and dis-tinction between universal and nonuniversal behaviorof magnetic overlayer systems when the coverage isfractional.26

D. Transport properties in magnetic systems

The study of transport properties in magnetic sys-tems differs from that in any other material by the factthat it always takes place in the presence of an intrin-sic, local magnetic field; in other words, it is always thestudy of galvanomagnetic properties—in particular,magnetoresistance.

When a magnetic field is applied to a normal (i.e.,not ferromagnetic) metal, the resistance is seen to in-crease with the intensity of the field, regardless of the

relative orientation of the field with respect to the cur-rent and with respect to the crystallographic axes. Thisphenomenon, known as ordinary or positive magneto-resistance, is very well understood, and for high-puritymetals with a large electronic mean-free path, yields ac-curate and easily interpretable information about theelectronic structure, the Fermi surface in particular, ofthe metal.27 Increases in resistance of many orders ofmagnitude (a factor of a million is fairly common) areobserved in particularly pure, single crystals at very low(liquid helium) temperatures and high magnetic fields(typically 10 to 100 kOe). For polycrystalline samplesand at normal temperatures more modest increases,typically of a factor of 2 to 10, are obtained for equiva-lent fields. Positive magnetoresistance can be inter-preted, in general terms, by noting that in the presenceof a magnetic field, electron trajectories become convo-luted (e.g., helical), and the effective distance that anelectron can transport charge before being scattered de-creases as the magnetic field increases.

In ferromagnetic systems, which in the absence ofan applied field consist of several magnetic domains,the phenomenon of negative magnetoresistance28 is ob-served: the application of an external magnetic fielddecreases the resistance by up to an order of magni-tude in fields as small as 100 Oe. The phenomenon iscommonly interpreted based on the fact that the ex-ternal field changes the domain structure and producesa single-domain crystal. Under those conditions twoeffects take place. The electron trajectories, because ofthe presence of a now uniform internal field, becomeless convoluted, and the removal of the Bloch wallseliminates a source of electron scattering.29 Both effectsresult in longer mean-free paths upon application ofa magnetic field, i.e. a negative magnetoresistance.

E. Micromagnetic theory

Micromagnetic theory provides a framework forpredicting macroscopic magnetic phenomena, such asdomain walls and hysteresis loops, in systems where thedetails of the atomic structure are not important.30

Input to the calculations includes exchange param-eters (typically taken from spin-wave dispersion data),crystalline anisotropy constants (typically taken fromtorque curves), and sample microstructure (typicallytaken from electron micrographs). It is a classical (i.e.,non quantum-mechanical) many-body problem in whichmuch of the computational expense comes from thelong-range nature of the magnetostatic interaction. Thememory dependence of the problem means that the mo-tion of the magnetization should be traced in timeto ensure accuracy: description by the Landau-Lifshitz-Gilbert equations appears to be adequate in this re-spect. The theory has spawned numerous calculationswhich, while usually only semi-quantitative in nature,


L.M. Falicov, D.T. Pierce et al.: Surface, interface, and thin-film magnetism

have provided considerable insight. Quantitative accu-racy is frequently prohibited by the need to include theeffects of thermal fluctuations and/or a precise domainnucleation mechanism as precipitated by defects. Fortu-nately, materials exist in which a nucleated domain canbe assumed to exist, and the major question in hys-teresis is whether the domain can pass some barrier.This approach led to a domain-wall pinning theory31

which predicted the approximate scaling of the coer-cive force with material parameters and to quantita-tively accurate predictions for hysteresis loops in CoNithin films.32 Quantitative accuracy for domain walls insoft materials (where details of microstructure andother complications may frequently be discarded) hasbeen achieved by several workers.33

III. MATERIALS

A. Growth techniques

In this section some commonly used preparationtechniques are described. The most extensively usedtechniques for the growth of modern magnetic materi-als are sputtering and Molecular Beam Epitaxy (MBE).Both techniques have produced high-quality samples,when grown under appropriate conditions.

1. Sputtering

In sputtering the target material is bombarded witha beam of inert gas ions (ordinarily argon) and the sput-tered atoms are collected on a temperature controlledsubstrate. Sputtering using magnetically confined plas-mas is ordinarily denoted by magnetron sputtering. Thepressure and substrate-target distance control the en-ergy distribution of the particles arriving at the sub-strate. Under appropriate conditions34 it is possible todeposit particles with an effective energy approachingthe sputtering-gas kinetic energy (—200 °C). Figure 1shows a comparison of the energy of Cu atoms arrivingat a substrate under typical evaporation conditions andunder the sputtering conditions specified in the cap-tion. Note that under these conditions, evaporatedatoms exhibit a high energy tail and are centered atmuch higher energies than sputtered atoms. This fact,of course, implies that under high sputtering pressuresand large substrate target distances sputtering producesless damage than thermal evaporation. An additionaladvantage of sputtering is that the energy distributionof particles can be tuned to higher energies by decreas-ing the pressure and/or the substrate-to-target distance.However, contamination due to the inert gas and thepresence of impurities such as oxygen makes sputteringinappropriate for the growth of semiconductors wherethe presence of small amounts of impurities is known

IO.UUU

1 10,000o

o

5 5000i

u_

0

1 1 1

/v

/ \ Sputtered

s*~V-^Evaporated

Cu

-

-

J-—

0.2 0.4 0.6

ENERGY(eV)0.8 1.0

FIG. 1. Energy distribution of particle flux arriving at a substratefor sputtering at a pressure of 10 mT and substrate-to-target dis-tance of 6 cm, and for thermal evaporation.34

to affect severely the physical properties. Moreover,because of the presence of the inert gas it is not custom-ary to use in situ characterization techniques. Sputter-ing is the growth technique of choice in industrialapplications where large-area homogeneous films arerequired at reasonable cost. Laser ablation is anotherrelated, more recent entry in the arsenal of the thin-film fabrication methods. In this technique a targetclose to the exact (or near exact) stoichiometry of thefinal films is bombarded ("ablated") by a laser beam, tomove the material from the target to the final film. Themethod is particularly well suited to those materialswhich have widely different sputtering rates. It has beenvery successfully used for the growth of high-tempera-ture superconducting oxides.35

2. Molecular beam epitaxy (MBE)

In MBE a number of particle beams are preparedby thermal evaporation from Knudsen cells or electron-beam guns in ultrahigh vacuum (UHV), typically of10~u Torr. The evaporation rates are kept slow andcontrolled using quartz crystal monitors, optical detec-tion methods, or mass spectrometers. Control of thesubstrate temperature and growth rate is essentialif smooth ledge growth is to be achieved, with sharpinterfaces and minimal interdiffusion. It is customaryto use in situ characterization tools such as ReflectionHigh Energy Electron Diffraction (RHEED) and inten-sity oscillation of the elastically reflected or diffractedelectron beams. The interpretation of the RHEED in-tensities and diffraction patterns has undergone a con-siderable evolution, although to date this techniqueis not on the same quantitative footing as x-ray or neu-tron diffraction techniques (see Sec. IV, Techniquesand Facilities).



3. Metal-organic chemical vapor deposition(MOCVD)

A variety of other techniques have also been usedto grow magnetic materials, especially oxides. How-ever, they have not been used as extensively as MBEand sputtering. MOCVD is a technique by which metalatoms are carried by a large easily dissociated organicmolecule from a source container to the substrate. Onstriking the substrate, the molecule momentarily sticks,but can be readily dissociated either by maintaining thesubstrate at a high temperature or by irradiating it withsufficiently-high-energy light. Common molecular car-riers are the metal carbonyls. Using iron pentacarbonyl,excellent single-crystal films of Fe have been grownepitaxially on GaAs at 175 °C substrate temperature.Post-growth analysis exhibited normal bulk magneticproperties and showed no evidence of entrapped car-bon.36 Plasma-assisted MOCVD takes advantage of theadded parametric control of the composition throughthe use of a "plasma" in the deposition chamber. Finetuning of the final composition occurs by adjustingboth the plasma and the source conditions for the de-sired result.

4. Production techniques

State-of-the-art systems typically might use dc-magnetron sputtering techniques with targets compara-ble to sample or sample pallet in size (a few inches indiameter for single samples to a couple of feet across forpallet systems). Deposition rates are kept high in large-volume products, if possible, in order to utilize effec-tively an expensive machine. Metals are commonlydeposited at rates as great as 200 A/s, at base pressuresapproaching 10"7 Torr. Substrates for different materi-als may require either heating or cooling. Amorphousmagneto-optic media on plastic substrates must obvi-ously not get hot. On the other hand, some crystalstructures, grain sizes, and material phases require sub-strate temperatures of a few hundred degrees Celsius(100-400 °C for magnetic media). Typically, multiplelayers must be deposited, so it is not uncommon to havein-line deposition systems. This fact requires compli-cated mechanical-transport systems if continuous pro-cessing, as opposed to batch processing, is to be used.Large volumes of deposited materials tend to flake offthe insides of vacuum chambers and create defects;hence most in-line systems are oriented and depositmaterial horizontally or with the substrate facing down.

Simple Chemical Vapor Deposition methods can beused to deposit a variety of oxides.37 The method essen-tially involves holding an appropriate single-crystal sub-strate close to a sublimating metal-halide source in thepresence of a pressure of about 15 mm of water vapor.Typical substrate temperatures are 700 °C, with bro-

mides used as cation. In this fashion thicknesses up toabout —20 /xm can be readily achieved.

Other processes may use electron-beam evapora-tion if very high rates are demanded for extended peri-ods of time. An example of this might be continuouscoating of a flexible substrate for videotape. In this casemaintaining material composition from the melt maybecome an issue and methods of monitoring compo-sition, and replenishing depleting alloy elements arerequired.

B. Growth modes

Generally it is hoped that the growth of a perfect,defect free, flat, completely segregated film can beachieved over large macroscopic areas of the substrate.In practice, the growth of films proceeds by a variety ofso-called "growth modes".38 These have been tradition-ally categorized in the epitaxial literature as the layer-by-layer, layer-and-island, and island growth modes.The particular growth mode depends on the relativebinding energies of the overlayer-overlayer atoms andthe overlayer-substrate atoms. In the case of heteroepi-taxial growth there are two limits: close to matching ofatomic structures of the overlayer and substrate ("lat-tice matching") and completely different structures andatomic radii ("lattice mismatched") systems. A hetero-epitaxial system which is close to lattice matching inmany cases will slightly strain (if the energetics permitsit) to match the substrate ("pseudomorphic growth").This strain can be partially relieved by the formation ofdislocations in thin films. Strained-layer properties anddislocation formation are current topics of investigationin semiconductor heterostructures, but have not beenfully addressed for the metallic systems that are themain topic of this report. Despite strain relief caused byplastic deformation at the growth temperature, a filmremains clamped to the substrate during subsequentmeasurements at reduced temperatures. These epitaxialconstraints can exert profound effects on the magneticphase diagram and on the general behavior of a partic-ular phase.39 For systems that exhibit large lattice mis-matches the substrate can predetermine the relativeepitaxial orientations and even the structure of theoverlayer through mechanisms which are not wellunderstood. Although a variety of epitaxial systems hasbeen grown over many years,40 many systems have beenprepared under uncontrolled conditions or poor vac-uum, making it unclear how much of the pre-existingliterature is relevant for issues concerning growth andstructure at the atomic level. Because structural andmagnetism issues are intimately related to each other,the understanding of the magnetism strongly dependson a detailed understanding of the structural properties.

Chemical and structural disorder at growth inter-faces are important in the overall issue of structure and


L. M. Falicov, D.T. Pierce etai: Surface, interface, and thin-film magnetism

magnetism. Structural disorder (roughness) can in-crease or decrease with growing thickness, and chemi-cal disorder depends on a variety of growth parameterssuch as substrate temperature, growth rate, energy ofdeposited particles, etc. Although growth is a nonequi-librium phenomenon and therefore kinetically con-trolled, thermodynamic phase diagrams are often foundto play a dominant role.40 The main reason is that sur-face diffusion may suffice to cause interdiffusion at alevel of 2 to 5 atomic planes at an interface betweentwo materials that form continuous sets of solid solu-tions. Indeed, materials that are lattice matched andhave the same crystal structure frequently form solidsolutions in their thermodynamic phase diagram.41 Allthese considerations point to the fact that it is veryimportant to characterize properly the materials afterthey are grown by a number of in situ and ex situ tech-niques (see Sec. IV, Techniques and Facilities). Thephysical properties may even reflect structural featureswhich are not readily detected by purely structuralprobes. One obvious example is the dependence of theelectrical resistivity on layer thickness in superlat-tices.42 It is generally found that the electrical resistivityscales inversely with the layer thickness, indicating thepresence of a large amount of electronic scattering atseemingly perfect interfaces.

C. Systems highlights

A large number of different systems have beengrown by the techniques described above (see Tables I,II, and III). These include single epitaxial films of al-

TABLE I. Lattice constant (in A) of representative substrate/filmcombinations for some magnetic metal films (for further details seeRef. 72).

TABLE II. Superlattice systems (for further details see Ref. 42,p. 139).

Substrate Film

fee

Cu (3.61)

fee

fee

Ni (3.62)Co (3.55)Fe (3.59)

bec

NaCl (5.64)AlAs (5.62)GaAs (5.65)Ge (5.66)ZnSe (5.67)

fee

Fe (2.867)x 2

5.733

bec

LiF (4.02)Al (4.05)Au (4.07)MgO (4.31)NaF (4.62)

Fe (2.867)x v5

4.054

System

Ni/CuNi/MoNi/CrNi/CNi/VNiFe/TiNCo/CuCo/AuCo/NbCo/SbCo/PCo/PdCo/CrCo/MnCo/GdCoNb/CoTiCoSiBi/CoTiFe/CuFe/AgFe/AuFe/SbFe/Sn

Preparation method

Ev, SpDCSpDCEvEvSpDCSpDCSpRF, EvSpRFEvEvElSpRF, EvEvEvSpDCSpSpSpRF, SpEvSpRFEvEv

System

Fe/MgFe/VFe/WFe/TaFe/YFe/PdFe/CrFe/MnFe/FeOFe/NdFe/GdFe/TbFeB/AgFeCo/SiFeCo/TbMn/AgMn/SbDy/YEr/YGd/YTm/Lu

Preparation method

EvEvSpDCSpDCEvEv, SpDCEvEvSpSpDCEvSpSpDCSpEvEvEvEv, MBEMBEEv, MBESp

Ev = EvaporationSp = SputteringEl = Electrolytic methodMBE = Molecular Beam Epitaxy

most all magnetic elements, including transition metalsand rare earths. Many of these elements have beengrown in ultrahigh vacuum down to submonolayerthicknesses.

TABLE III. Metal-on-metal growth (for further details see Ref. 73).

Overlayer

Substrate Cr Mn Fe Co Ni Rare earth

Cu(100)Cu(110)Cu(lll)Ag(100)Au(100)Pd(100)Pd(lll)Ru(0001)Ru(1010)Re(0001)W(UO)

X

X

X

X

XXXXXXXXX

X

X

X

XX

Fe(100)Ni(100)CuAu(lll)Cu3Au(100)Y(0001)

X

XX

Eu Gd TbCe Dy

Gd

Dy Er Gd Ho


L. M. Falicov, D.T. Pierce ef a/.: Surface, interface, and thin-film magnetism

1. Surface and monolayer films

While it is possible for theorists to model idealmonolayers in computer simulations, it has proven analmost insurmountable challenge for experimentaliststo grow idealized model systems in the laboratory. Theissue is associated with the need for a substrate and theinability to realize free-standing monolayers. Inter-actions with the substrate invariably dominate mostproperties of interest. For instance, Cu, Ag, and Ausingle crystals are good substrate candidates because oftheir filled d bands. But it is this very characteristic thatcreates metallurgical problems: their lower surface freeenergies, compared to that of the magnetic elements,can provide thermodynamic driving forces for surfacesegregation, intermixing, etc. For the Fe/Cu(100) systemit is known that intermixing at the interface preventsthe realization of monolayer structures.43'44 Elevatedsubstrate temperatures during growth of multilayerthicknesses of Fe on Cu(100) have been shown to providean intermixed buffer layer that separates the pure Feand pure Cu regions; this buffer layer stabilizes theantiferromagnetic phase45 of fee Fe. On the other hand,the relatively discrete interface formed by low growthtemperatures yields a ferromagnetic fee phase of Fe.46'47

This is a clear example of the influence of growth con-ditions on the properties of the resultant film.

Transition-metal substrates should have higher sur-face free energies. However, hybridization between themagnetic d or / electron states and the substrate d elec-tron states across the interface becomes a controllingfactor. For instance, it has been shown that while feeFe(lll) grows on Ru(0001) with an expanded in-planelattice spacing that should promote ferromagnetism, thefirst two monolayers of Fe appear to be magneticallydead.48 The explanation is that the in-plane expansionleads to an interplanar contraction and a strong Fe-Ruband hybridization that precludes magnetic momentformation. For the Fe/Pd(100) system, quite the oppo-site effect occurs. The strong d-d hybridization is pre-dicted to induce ferromagnetism in the Pd substrate.49

It is interesting to note that these trends are mirrored inthe behavior of dilute Fe alloys in 4c/-transition-metalhosts: Fe in Ru lacks a local moment, while Fe in Pd isthe classic giant-moment system because of the polari-zation of Pd sites that extends many atomic shells awayfrom the impurity site.

The structural, morphological, and growth-modecorrelations with magnetic properties present an on-going challenge to materials researchers workingwith monolayer and ultrathin magnetic-film structures.It is well documented for the Fe/Cu(100) system thatgrowth-temperature and film-thickness variationschange the magnetic spin orientations in a systematicmanner.46 A magnetic anisotropy diagram has beenconstructed (Fig. 2) to summarize the results. Subse-

1

300

200 -

15

100 " • •

JT

• • • • OS S S

0 2 4 6 8

Thickness (ML)

FIG. 2. The region of stability of perpendicular anisotropy for feeFe/Cu(100) outlined on a plot of growth temperature versus filmthickness in monolayers (ML). The Kerr-effect measurements usedto determine the stability boundaries were made at the growthtemperature.46

quent studies48 suggest that there is a degree of gener-ality to the systematics observed in Fig. 2. However,the detailed structural underpinnings of the observedtrends remain elusive.

The perfection of surface structures impacts on thestudy of critical phenomena in two dimensions as well.Imperfections can prevent the magnetic correlationlength from diverging as the critical temperature is ap-proached. This broadens the transition and couples thedata-analysis task of defining the magnetization expo-nent to that of simultaneously defining an effective Tc.

Surface perfection also manifests itself in the questto verify the theoretical predictions regarding possibleferromagnetism50"52 at the {100} surfaces of Cr. Thisprediction also indicates that the moments are dramati-cally enhanced at the surface. The surface-orderingtemperature is also raised, relative to the Neel temper-ature of bulk Cr. The enhanced surface magnetism ofCr(100) leads to ferromagnetic (100) sheets that are cou-pled antiparallel to each other on adjacent layers.50"52

The problem is that if terrace widths at the surface aresmaller than the domain-wall thickness, the surface be-comes divided into antiparallel domains, and there isno net moment on a macroscopic scale. Since even a(100) surface well-defined by standard surface-sciencecriteria does have step densities of order one per 100 A,even with polarized-electron imaging of the domainstructure the present resolution level (—500 A) is insuf-ficient to clarify this exciting issue. The future shouldbring increased experimental resolution in imaging andan enhanced ability to create ultraflat surfaces, e.g., byepitaxy on GaAs-based heterostructures, or via homo-epitaxial smoothing of the surface as part of the fab-rication process.



2. Metastable epitaxial films

Elemental magnetic materials exist in a varietyof crystallographic and magnetic phases. Thin-filmgrowth of these materials on crystalline substrates al-lows the forces present at the interface to drive thefilm into specific crystallographic structures. Thesestructures may be a known high-pressure or high-temperature phase, or a phase not previously observed.Since the energies associated with a change in crystalstructure (—0.1 eV per atom) are of the same order ofmagnitude as energies associated with a change in mag-netic structure (e.g., ferromagnetic to antiferromag-netic), often the magnetic properties of thin filmsdramatically depend on the growth conditions andstructure of the substrates. These artificial magneticmaterials, which are stabilized by their growth in thinfilm form, are collectively referred to as metastablestructures and now form the basis of an active field ofresearch. They greatly expand the number and varietyof magnetic materials by essentially making new mate-rials from "old" elements.

In addition to providing new structures, thesemetastable phases provide stringent tests of calculationaltechniques used to predict structural and magneticproperties of magnetic materials. These techniques (seeSec. II, Theoretical Background) are capable of yieldingthe total energy of an elemental crystallographic systemas a function of lattice structure and spacing, includinga zero-temperature prediction of magnetic moment andmagnetic structure. Indeed, the failures of current cal-culations to predict accurately the energy hierarchies ofthese phases are helping to pinpoint the deficiencies inthe theoretical underpinnings.

An example of the richness of phases available to amagnetic element is provided by inspecting the phasediagram54 of Fe shown in Fig. 3. At ambient pressureand temperature the common bcc form of ferromagneticiron is obtained. At high pressure and low temperature,however, the hep e-phase, which is nonmagnetic, is pre-dicted. This is the expected phase in the absence ofmagnetic effects, as given by the structure of the otherelements in the iso-electronic sequence, Ru and Os.At higher temperatures paramagnetic fee y-Fe and bccS-Fe are predicted, and at even higher temperaturesthe system melts.

The total energies calculated55 for the cubic phasesare shown in the top frame of Fig. 4. These calculationsshow a clear energy minimum for bcc Fe at the ob-served lattice constant and correctly predict it to beferromagnetic. A nonmagnetic fee phase is predictedfor a smaller Wigner-Seitz radius at slightly higherenergy and a second ferromagnetic fee phase at a largerWigner-Seitz radius at much higher energy. (These calcu-lations were based on spherical approximations to theatomic potentials and charge densities; more accurate

1500 r

ICCUJ0 -

UJ

1000 -

5 0 0 -

. (bcc)

-

i > '

a(bcc)

i i

7(fee)

1 1 1 1

\ (hep)

i i \ i

" 0 50 100

PRESSURE (kbar)

FIG. 3. Phase diagram for bulk Fe.54

nonspherical calculations yield, erroneously, the feephase as the most stable one for iron.9) The point atwhich these two branches cross corresponds to a latticeconstant very close to that of fee Cu (bottom frame ofFig. 4). Epitaxial growth of Fe on a Cu substrate hasshown that either ferromagnetic or antiferromagneticfee Fe can be obtained, depending on the detailed con-ditions of growth (substrate temperature, surface prepa-ration, and surface cleanliness).45 This indicates thateven fine details of total energy calculations may bemanifest in metastable thin films.

Another example of a metastable phase is given inthe second frame of Fig. 4. It shows total energy curvesfor two cubic phases of Co. Face-centered cubic Co isthe high temperature ferromagnetic phase observed innature; however, there is no naturally occurring bccphase of Co. Experimentally, however, a bcc ferromag-netic phase53 was successfully formed by epitaxialgrowth on GaAs. Total-energy calculations55 yield thebcc-Co phase with the observed lattice constant, shownin Fig. 4, and correctly predict it to be ferromagnetic.

Finally, Fig. 4 indicates that there should be bccphases of Ni, both ferromagnetic and nonmagnetic,even though in nature Ni appears only in a ferromag-netic fee phase. Body-centered cubic Ni has been re-ported to be stabilized by epitaxial growth on a singlecrystal surface of Fe(100). At this lattice constant, itis far from the metastable equilibrium value for theWigner-Seitz radius indicated by the calculation, and



1.30 1.35 1.40 1.45

FIG. 4. Calculated total energy versus Wigner-Seitz cell radius forseveral metals.55

the strong influence of the ferromagnetic substratemade magnetic characterization difficult.

3. Semiconductor substrates

Single-crystal semiconductor substrates provide avery attractive template for the epitaxial growth ofmetal films. In particular, a group consisting of Ge,GaAs, AlAs, and ZnSe all have lattice constants veryclose to 5.65 A. This is also very close to twice thelattice constant of bcc Co (2.82 A), bcc Fe (2.87 A), andbcc Ni (2.89 A), which should permit a c (2 x 2) recon-struction of the metal films upon these substrates. Al-

though bcc Co has been successfully grown on GaAs53

and bcc Fe on Ge, GaAs, and ZnSe,56 there is an impor-tant issue of interface chemistry with these systems. Ithas been observed, for example, that although Fe filmsgrown on Ge have excellent structural quality, there isconsiderable interdiffusion at the interface, which di-minishes the magnetic moment. When Fe is grown onGaAs, photoemission studies57 show that FeAs isformed at the interface, releasing a partial monolayerof Ga which is then covered by subsequent depositionof Fe. Commercially processed substrates, however,can release significant amounts of As which largely dif-fuse to the top surface of the Fe film during growth.Furthermore, the small amounts of As incorporated inthe film within an exponential decay length of 10 Afrom the surface show an inordinately large effect indiminishing the moment up to 100 A from the inter-face. These effects can be eliminated by first cappingthe substrate with a homoepitaxial layer of GaAs or anepitaxial layer of ZnSe. The growth of Fe on ZnSeepilayers shows the full Fe moment in films down to20 A thickness and the cubic anisotropy of bulk Fe.

4. Rare earths

The growth of rare earths provides a particularlyfertile ground for the study of magnetic phenomena inthin films and their relationship to magnetism in re-duced dimensionality. The main reason is that rareearths display a variety of systems which are chemicallysimilar, span a large range of ionic radii and crystalstructures, and present a wealth of magnetic structuresincluding helical, ferromagnetic, antiferromagnetic,and cone magnetic structures. In addition, rare earthsexhibit a great variability of thermodynamic phase dia-grams ranging from complete immiscibility—as is thecase for many rare earths with transition metals—tothe formation of complete sets of solid solutions—as isthe case of two rare earths. The epitaxial growth ofrare earths and transition metals is particularly chal-lenging because of the high reactivity of the rare earthsand the high melting points of many of the transitionmetals. As a consequence, MBE is used for these sys-tems, with special care taken to avoid contamination.Of course, as in all MBE growth, the structure of theepitaxial layer is monitored in situ using RHEED,RHEED oscillations, and ex situ x-ray and neutron dif-fraction. Generally it has been found that the growth ofrare earths can be accomplished quite conveniently ona transition metal, for instance Gd on W(110),58 Y onNb(llO),59 or Ce on V(110).60 One reason that these sys-tems can be grown with relative ease is that they do notform solid solutions in their phase diagram, and possi-bly this facilitates the growth of a segregated rare earth.

The growth of epitaxial rare-earth films and multi-layers had as a key ingredient the discovery59 that



Y(0001) grows epitaxially on Nb(llO). Following thegrowth of 50-100 nm of Y(0001), excellent rare-earthfilms and multilayers can be produced. It is currentlypossible to produce films and multilayers whose crys-talline perfection, as measured by Bragg peak widthand mosaic spread, rivals that of bulk single crystals ofthe same elements. The elements Gd, Dy, Er, and Hohave been extensively studied, as have multilayers ofthe same elements, separated by yttrium, rare-earth-yttrium alloys, or other rare earths and alloys.39'41'61"63 Arecent development in this field uses homoepitaxy toproduce a chemically clean surface on a substrate mate-rial, which can have any desired crystallographic orien-tation. For instance, crystals of hexagonal Y are cut toexpose {1010} and {1120} faces after which additional Yis grown epitaxially. These have then been used as sub-strates on which to grow {1010} and {1120} rare-earthfilms and multilayers.64 Such samples are important forthe study of the influence of epitaxial constraints onmagnetic properties, and open a new field for the studyof propagation of magnetic order along different crys-tallographic directions.65 This technique has also beenused to produce rare-earth films and superlattices onthin substrates. Niobium foils were grown as (110) singlecrystals (1 cm2 x 5 fim),M followed by the same proce-dure used to grow rare earths on sapphire substrates.Such samples are particularly useful for x-ray scatteringstudies in transmission geometry, for mechanical andthermodynamics measurements, and for electron mi-croscopy of epitaxial samples.

The epitaxy of rare earths on vanadium has beenpursued in order to understand the role that latticematching and chemistry play in the growth.60 This sys-tem exhibits complete immiscibility with most rareearths and therefore it was possible to study how itsgrowth is affected by lattice mismatch. In particular,Ce(lll) on V(110) has shown the presence of a new epi-taxial orientation, not yet observed nor predicted in anyfcc(lll)/bcc(110) system,66 and the stabilization of anew metastable phase of Ce, expanded in the directionsparallel and perpendicular to the growth direction. Thegrowth of dysprosium on vanadium exhibits a variety ofnovel expanded phases for dysprosium as well as a se-ries of surface reconstructions as a function of thick-ness.67 It is quite interesting to note that the expandedphases are not governed by the Poisson's ratios of theoverlayer, and that the expansion occurs in all direc-tions, similarly to the earlier observations for thegrowth of Fe on GaAs, where a contraction occurs inall directions.

5. Oxides

One particular type of system which is of great im-portance and which has not been studied extensively isthe growth of epitaxial oxides. Oxides in many cases

exhibit interesting magnetic properties—such as anti-ferromagnetism—and are the basis for a variety ofdevices, especially when used in conjunction with aferromagnetic material. The growth of oxides has alsoreceived enhanced notoriety because of the discoveryof high-temperature superconductivity in ceramic ox-ides.68 The growth has usually been accomplished usingoxygen sources in an MBE system, using reactive sput-tering or laser ablation techniques. Nickel monoxide(NiO) and cobalt monoxide (CoO) single crystals havebeen prepared on MgO substrates using CVD.69 Re-cently titanium oxides were grown on sapphire by MBEusing activated oxygen sources.70 Chemical vapor depo-sition was used to prepare a variety of thick oxide films,especially ferromagnetic compounds such as NiO, CoO,Ni^Coi-jO, and RFeO3 (where R is a rare earth).71

6. Multilayers

A large variety of multilayered systems have beengrown: ferromagnetic-normal metals, ferromagnetic-superconducting, rare-earth-rare-earths, etc. The pre-ferred growth method has been sputtering of MBE, al-though recently titanium-oxide-titanium superlatticeshave been grown by the CVD techniques describedearlier.

Multilayered systems which are lattice-matchedhave been grown by thermal evaporation or MBE. TheMBE grown, lattice-matched systems exhibit narrowx-ray diffraction lines comparable to the instrumentalresolution. The lattice mismatched systems are gener-ally textured and exhibit broader x-ray diffraction lines.However, questions regarding interfacial chemistryhave not been fully addressed, because detailed under-standing of roughness, disorder, and interdiffusion isonly now being addressed (see Sec. IV, Techniques andFacilities).

D. Physical properties

1. Proximity and interfacial effects

In some systems, interface effects of a purely mag-netic origin extend beyond the interface and into thebulk, thus giving rise to a proximity effect. Examplescan be found in transition-metal systems where one sideconsists of a strong ferromagnet, such as Co, and theother side consists of an easily polarizable (almostmagnetic) material, such as Pd, or a weakly magneticmaterial, such as Cr. The strong electron-electron inter-action of the fully saturated ferromagnet, frustratedby a lack of d holes from producing a larger moment,induces through hybridization and exchange an addi-tional magnetic moment in the d bands of the polarizablematerial. This effect is analogous to the polarization ofthe Fe atoms in dilute Fe-Co alloys74 and the polariza-tion of Pd atoms in dilute Pd-Fe alloys.


L.M. Falicov, D.T. Pierce et al.\ Surface, interface, and thin-film magnetism

Theoretical and experimental studies of proximityeffects in transition and simple metals have establisheda series of empirical rules that can be summarized asfollows75"77:

(1) The magnetic moments of cobalt and nickel arevirtually saturated; they can be only very slightlychanged by their immediate environment. The frac-tional change, however, can be appreciable in nickel(which has a small moment of about 0.6 Bohr magne-tons), but is negligible in cobalt.

(2) The magnetic moment of iron, which has only amoderate electron-electron interaction, can be appre-ciably affected by its immediate environment.

(3) Chromium, which is a weak magnetic ion, mayhave its moment profoundly altered by the presenceof surfaces, interfaces, and both magnetic and non-magnetic neighbors.

(4) The "almost magnetic" elements, vanadium andpalladium, may acquire a sizable magnetic moment inthe proper environment.

(5) Free surfaces, which reduce the local band-width of a metal, tend to increase the magnetic momentof an element; hence the surface of chromium has amuch larger moment than the bulk,78-79 nickel tends tobe marginally more magnetic at the surface,80 and it ispossible that some crystallographic faces of vanadiumexhibit a magnetic moment.75

(6) Proximity of a nonmagnetic metal tends tosuppress the magnetic moment of some elements; thiseffect depends crucially on the overlap of the wavefunctions between the d band of the magnetic metaland the conduction band of the nonmagnetic one.

(7) The proximity of a strongly magnetic elementtends to induce or enhance magnetic moments on theneighboring, susceptible elements. Thus iron becomesmore magnetic in the proximity of cobalt,74'81 theenhanced moment of the chromium surface tends topropagate over several layers into the bulk,82 chro-mium acquires a large moment in the proximity of iron78

and/or cobalt, and vanadium and palladium may de-velop sizable magnetic moments in the proximity ofiron and/or cobalt.

2. Exchange coupling across interfaces

Magnetic exchange coupling between ferromagneticand antiferromagnetic layers was originally discoveredin oxidized Co particles.83 The antiferromagnetic CoOsurface layer is exchange-coupled to the ferromagneticCo interior which results in an imposed unidirectionalanisotropy. As a result, an asymmetric hysteresis loopshifted from zero field by the exchange bias fieldHb developed. A second thin-film system which hasbeen extensively studied84 is ferromagnetic permalloy,Ni8iFe19, coupled to the antiferromagnetic alloy, FeMn.For a ferromagnetic/antiferromagnetic coupled system

to exhibit a macroscopic exchange bias field Hb, theNeel temperature of the antiferromagnet must be lowerthan the Curie temperature of the ferromagnet.85 In ad-dition, the magnetic anisotropy energy of the antiferro-magnet must be large compared to the interfacialexchange coupling so that the antiferromagnetic spinsystem remains substantially blocked when the magne-tization of the ferromagnet rotates upon application ofan external field.

The magnitude of the interfacial exchange cou-pling energy, Ett, is much larger in the Co/CoO systemthan in NisiFeu/Fe^Mrii-*. In the latter case the mag-nitude of Ea is approximately 100 times smaller thanexpected in the simplest model: an antiferromagneticstructure comprised of uncompensated ferromagneticlayers whose magnetization, directed normal to the in-terfaces, alternates in sign from layer to layer.86 Assum-ing no relaxation of the antiferromagnetic structure atthe interface, Hb is given by (J/M), where / is the inter-facial exchange coupling and M is the magnetization ofthe ferromagnet. Values of/ comparable to those in theferromagnet or the antiferromagnet yield, for Hb, val-ues 100 to 1000 times greater than those observed ex-perimentally. The experimental data are surprisinglyconsistent for several NigiFeig/Fe^Mni-* films pre-pared by a number of different workers under quite dif-ferent conditions.

Various models have been proposed to account forthe large discrepancy between experiment and theorynoted above. However, none of them can explain allthe properties of this coupled system. It is quite clearthat no domain wall is formed in the ferromagneticlayer via some sort of "wetting" to the antiferromagneticlayer upon rotation of its magnetization. The possibilityof a planar domain wall in the antiferromagnetic layercan be ruled out, since antiferromagnetic layer thick-nesses thinner than a typical domain wall by an orderof magnitude give the same Hb. The most completemodel proposed so far86 suggests that interfacial atomicroughness would reduce the exchange coupling energy.Therefore roughness or chemical inhomogeneities atthe interface are assumed to give rise to random inter-facial exchange interactions (parallel or antiparallel tothe direction of the unidirectional anisotropy). The an-tiferromagnet minimizes its energy by breaking up intolateral domains whose size is approximately that of theantiferromagnetic domain-wall width. Thus, averagingover the random exchange fields in a single domain im-plies that the interfacial exchange coupling energy isreduced by Nm, where N is the number of atoms in theinterface layer of a single antiferromagnetic domain.This model thus predicts that the larger the antiferro-magnetic domains, the greater the extent to which therandom fields cancel one another. The observed valuesof Hb in NigiFew/MrijFei--* are reasonably consistent


L.M. Falicov, D.T. Pierce etai: Surface, interface, and thin-film magnetism

with the model. Further refinements, including a moredetailed description of the antiferromagnetic structureof MnFe, are an important prerequisite for improvedunderstanding of this exchanged coupled system. Thedetails of the interfacial structure and their effect onmagnetism are a crucial part of this understanding.

Not only do ferromagnetic/antiferromagnetic cou-pled systems display a fascinating range of properties,but the interfacial exchange coupling can be harnessedto study the properties of the antiferromagnetic layer. Itis extremely difficult to measure many fundamentalmagnetic properties of ultra-thin antiferromagneticfilms, including, for example, their Neel temperatures,because of the difficulty of coupling to the sublatticemagnetization. Most electron, optical and neutron scat-tering, and magnetic resonance techniques are in-capable of examining antiferromagnetic thin films.Spin-polarized photoelectron diffraction is one of thefew techniques with some potential for such studies.This technique, however, is in its infancy and is sensi-tive only to the magnetic short-range order which, forthe two antiferromagnetic single-crystal systems so farstudied,87 persists to temperatures several times largerthan the bulk Neel temperature.

The ferromagnetic layer in a ferromagnet/antiferro-magnet couple forms a natural probe of the antiferro-magnetic system. By monitoring the temperature atwhich the exchange bias field goes to zero, the blockingtemperature of the antiferromagnet can be determined.This temperature is slightly lower and closely related tothe Neel temperature TN of the antiferromagnet. At atemperature just below JN the anisotropy of the anti-ferromagnetic layer becomes too weak compared to theexchange coupling energy to maintain the rigidity ofthe antiferromagnetic lattice, which thus becomes freeto follow the magnetization of the ferromagnetic layer.The dependence of the blocking temperature on thethickness of the FeMn layer in polycrystalline filmsof NisiFeig/FctMni-* has been determined88 for aconstant NiFe thickness of about 60 A. The blockingtemperature is independent of thickness for FeMnthicknesses greater than about 100 A, but it is lower forthinner layers. The thickness dependence has beenfound to follow a simple finite-size scaling relationship.This method can clearly be applied to other antiferro-magnetic systems and should prove to be a rich area forfurther work.

In contrast to the ferromagnetic/antiferromagneticcoupled systems, the magnitude of the exchange cou-pling in ferromagnetic/ferromagnetic systems can bevery large. A wide variety of systems has been studied;they include, however, very few studies on well charac-terized single crystals. One example of the latter aresingle crystals of bcc Ni/Fe bilayers.89 For Ni layerthicknesses greater than six monolayers the Ni lattice

reconstructs and, via exchange coupling to the Fe layer,imposes a large in-plane fourfold anisotropy on the Felayer. Less well characterized systems include polycrys-talline Ni81Fei9/Fe superlattices and a wide variety ofamorphous rare earth-transition metal (RE/TM) alloyfilms coupled to other RE/TM alloys or polycrystallinefilms of Fe, Co, or Ni8iFei9. The latter systems all havebeen developed for their possible application in a vari-ety of magnetic recording devices. Superlattices ofNi81Fei9/Fe with layer thicknesses in the range, forNigiFew and Fe respectively, of =100-500 A and =300-1500 A combine the high saturation magnetization withthe high permeabilities required for magnetic record-ing-head applications.90 Whereas in these superlatticesthe saturation magnetization is simply the appropriateaveraged saturation of the Ni8iFei9 and Fe layers, thecoercivity of the superlattice is much closer to that ofNi8iFei9 than to that of Fe.

There is a large body of work on exchanged-coupledRE/TM systems. Spin-polarized photoemission studieson Fe/TbFe have shown that thin Fe layers take up theperpendicular anisotropy of the amorphous TbFe un-derlayer, with hysteresis loops which reflect strong ex-change coupling between layers.91 Exchange couplingbetween Ni8iFei9 and TbCo leads to exchange-shiftedhysteresis loops with bias fields92 as large as 500 Oe forNi8iFe19 layer thicknesses of about 400 A. A variety ofschemes for taking advantage of the magnetic exchangecoupling between two different RE/TM alloys has beenproposed93; these schemes optimize the performance ofa magneto-optic storage medium. In particular, oneof the RE/TM layers is chosen to have a high magneto-optic rotation, 6K, in the wavelength range of interestwhereas the second layer, which may have a small 6K, ischosen to have a high coercivity. Thus the first layerhas optimum read-out properties and the second, opti-mum storage properties.

3. Coupling through nonmagnetic layers

One of the most interesting ferromagnetic/metal/ferromagnetic systems is Fe/Cr/Fe, where it has beenfound that successive Fe layers, for thin Cr layer thick-nesses, are coupled antiparallel to each other. The cou-pling diminishes as the Cr layer increases in thickness.The coupling mechanism is, at present, not yet wellunderstood. It is apparently too large a coupling to beaccounted for by magnetostatic effects. Such an anti-parallel coupling was observed in Fe/Cr/Fe trilayerstructures94 in spin-polarized electron-scattering studiesand in Magneto-optic-Kerr-effect (MOKE) and Brillouinscattering studies.95

Recently the same phenomenon has been observedin MBE grown Fe/Cr superlattices by magnetizationstudies. The superlattice has no moment in zero field,



but the moments of the antiparallel, neighboring Fe lay-ers can be aligned by application of fields of up to20 kOe for Fe layer thicknesses of 30 A, which impliesvery large effective exchange-coupling energies. An im-portant property of these structures is the large drop inresistance observed on aligning the Fe layer moments.96

This "giant" magnetoresistance effect is not yet fullyunderstood; it is of great interest for potential recordinghead applications.

There have been numerous studies of systems ofthe ferromagnet/metal/ferromagnet type, ranging fromattempts to vary the coercivity of ferromagnetic filmsby lamination for magnetic recording applications, tostudies of single-crystal superlattices,97'98 such as Fe/Ag.Exchange coupling of successive Fe layers in this sys-tem has been inferred98 from the temperature depen-dence of the magnetization at low temperatures. Acalculation99100 of the temperature dependence of themagnetization in the spin-wave regime for an arbitrarymultilayered magnetic structure has shown that therealways exists a range of temperature for which the mag-netization varies as aT312, where the coefficient a de-pends on the exchange coupling between the magneticlayers. The method of calculation can be applied toobtain the exchange coupling in ferromagnetic/metal/ferromagnetic systems. In some recent elegant experi-ments101 the coupling between a surface layer of Ni81Fe19

and an underlying thick Ni8iFew layer (separated fromeach other by submonolayers of Ta) was obtained.

Tunneling between a spin-polarized superconduct-ing film coupled to a ferromagnetic layer has been ex-tensively used to study the magnetic properties of thinferromagnetic layers.102 It has been proposed103 thattunneling between two ferromagnets could depend onthe relative alignment of the magnetization of the twoferromagnetic layers; this effect was subsequently ob-served104 in the system Fe-Ge-Co. The magnitude ofthis magnetic tunneling-valve effect was found to beabout half that expected from the spin polarizationsin Fe and Co as deduced from tunneling105 in ferro-magnet/insulator/superconductor junctions. Morerecently, similar effects have been observed106 inNi/NiO/Co tunnel junctions.

4. Ruderman-Kittel-Kasuya-Yosida (RKKY) coupling

Bulk rare-earth elements and their alloys with yt-trium exhibit complex spin arrangements caused by thecombination of strong crystal field effects and the oscil-latory exchange interaction modulated by the conduc-tion electrons (RKKY interaction). Early work inrare-earth multilayers107108 demonstrated that RKKYpolarization propagates across the rare-earth/yttrium(0001) interface, and thus it decays slowly enough toprovide coherent exchange coupling across as much as130 A of Y. Spiral (transverse) and c-axis (longitudinal)

polarizations are preserved. Experimental results ofmagnetic neutron scattering, which provide evidencefor the propagation of magnetic order through the non-magnetic Y, are shown in Fig. 5. Intriguingly, the peri-odicity of the spin polarization in the Y is that of diluterare-earth-yttrium alloys, while that in the rare earthdeviates from bulk values at low temperatures.

A model for RKKY coupling has been proposed.109

Rare-earth sheets were required to be immersed in theY conduction band, but to interact with the s-f inter-actions appropriate to the rare earth. The polarization,therefore, is formed by the nesting features of theY band structure. In the case of spiral structures,two transverse polarization waves, out of phase by onelattice spacing, are produced, thus providing a helicalarrangement. A later refinement of this picture postu-lates the existence of a superlattice band, with the os-cillating polarization being a feature of the hybridband. Wave vector conservation parallel to the inter-face prevents mixing of states at the Fermi surface withdifferent interface momenta, and can result in local-ization of certain electron states on one side of theinterface.

Recent experiments have explored the RKKY cou-pling across (1010) and 1120) interfaces. For Dy/Y thepolarization is insufficient to bring the spiral order ofsuccessive rare-earth blocks into coherence but doesprovide adequate coupling to produce long-range ferro-magnetic order in Gd superlattices. This may simplyreflect the strongly anisotropic range of RKKY oscilla-tions64 in Y but may also be evidence for total reflectionof those conduction electrons most important in provid-ing the RKKY coupling, as presented by the super-lattice band approach of Ref. 109. Other evidence forspin-dependent transmission has been seen in magneto-transport experiments in the Fe-Cr-Fe system.

5. Magneto-elasticity

The presence of strain has been used to modify thephysical properties through the magnetoelastic effect.This is particularly important for materials such as rareearths and Laves-phase alloys, where magnetoelastic ef-fects are large. This effect was first discovered in Dysuperlattices39 and films,67 where the magnetoelasticallydriven ferromagnetic transition is suppressed. Similareffects have been observed in Er films and super-lattices. Both Dy and Er epitaxial materials can bedriven to ferromagnetism at a critical value of the ap-plied field that depends on film thickness. In the caseof Er, a variety of commensurable spin states are in-duced at low temperatures by fields below the criticalvalue.110 Bulk behavior is not recovered in films up to1 jam thick. The treatment of this problem to date hasrelied on bulk values of the magnetoelastic couplingconstants subject to rigid clamping assumptions. Mea-


L. M. Falicov, D.T. Pierce etal.: Surface, interface, and thin-film magnetism

100-

80-

60-

OO

40-

20-

(a)

Er film860 A

.4 0.0

A.

J I I L

10 r

65K A ,002"

(b)

101+

101

, M

002+

FIG. 5. Magnetic neutron scatteringfrom (a) an Er film, and (b) an Er/Ymultilayer with 32 atomic planes ofEr and 21 of Y. The multiple peakstructures near ±0.3 A"1 demonstratethe coherent propagation of magneticorder through the nonmagnetic yttriuminterlayers.110

0.4

Qz-rz (A"1)

surements of actual strains in epilayers, along with aproper treatment of magnetoelasticity for such systems,are clearly required.

A direct measurement of the RKKY coupling canbe made by studying the break-up of long-range coher-ence by applied fields at temperatures where magneto-elastic effects are most important. Because there is anet moment in incommensurable spiral layers, appliedfields align these moments, directly acting as a randomfield on the spiral. The loss of coherence in Dy/Y typi-cally occurs39 on application of fields of the order of5000 Oe.

6. Superlattice effects

Many of the effects described above can be conve-niently studied in multilayered films since they consistof a superposition of single films.111 Moreover, multi-layers provide the possibility of ex situ studies withoutconcern regarding contamination, since they can begrown very thick (~1 /um) compared to usual contami-nation depths. It should also be pointed out here thatmultilayers offer an additional dimension in characteri-zation, as described in Sec. IV (Techniques and Facili-ties). The drawback is, of course, that by its very natureany single, bi-, or tri-layered film effect can be obtained

only in a statistical sense, averaged over many (hope-fully equivalent) repetitions of the system.

There is, however, a class of effects which cannot,even in principle, be observed in a small number of lay-ers because they rely on the periodic nature of the multi-layer. These are the so-called superlattice effects. Theoriginal observation of this type of effect was that ofphonon folding in semiconductor superlattices.112 Inmetal systems there have been several unobserved the-oretical predictions of minigaps in the electrical trans-port phenomena,113 of localized states in these gaps, andof gaps in the continuum of the density of states insuperconductors.114 These effects all rely on the pres-ence of extended electronic states in the growth direc-tion. However, all metal systems studied to date exhibitlarge amounts of interfacial scattering, as indicatedby the thickness-dependent resistivity.42 Whether thisscattering is sufficient to break down the existenceof extended states perpendicular to the layers and ineffect confine the electrons to individual layers is notclear at this time. Possibly these effects should be ob-servable in high-perfection superlattice systems whichexhibit no interfacial scattering.

A superlattice effect which does not require perfec-tion at the atomic level is the development of the


L. M. Falicov, D.T. Pierce etai:. Surface, interface, and thin-film magnetism

magnon bands in ferromagnet/normal-metal super-lattices.115 The coupling in these types of superlatticesdepends on the long-range dipolar interaction whichis not much affected by small amounts of disorder atan interface. The individual modes in each one ofthe magnetic layers spreads into bands of magnons asthe intervening normal metal thickness is decreased.This is illustrated in Fig. 6 where a qualitative plot ofmagnon frequency versus normal-metal separation isshown. These predictions are conceptually similar tothe development of energy bands in a metal from thediscrete electronic levels present in individual atoms.Detailed theoretical predictions115 have been obtained,including the dependence of the magnon frequencies onlayer thicknesses, magnetic field, saturation magnetiza-tion, and wave vector. All these have been verified indetail in a series of measurements in Ni/Mo super-lattices.116 As an example, Fig. 7 displays the magnonfrequency as a function of magnetic field for a numberof superlattices. The excellent agreement between ex-periment (plus signs) and theory (solid line) shows thatour understanding of this phenomenon is on a firmfooting. It is even more striking that all parameters(thicknesses, magnetic field, magnetization, wave vec-tor) that enter into the calculations are independentlymeasured so no parameters need to be adjusted to bringtheory and experiment into agreement.

IV. TECHNIQUES AND FACILITIES

The new scientific opportunities in magnetic-materials research are ripe for exploration because ofthe modern research techniques and facilities available

BulkIsolated Thin

Film

Surface

Bulk

Surface

'Normal

FIG. 6. Qualitative development of magnon bands in a magnetic/normal-metal superlattice. The figure shows the magnon fre-quency as a function of normal-metal thickness tN; tM is themagnetic-metal thickness. The shaded area between the two linesrepresents the band of superlattice modes.115

to the materials science community. The techniquesspan the range from the structural-characterizationtools shared with the semiconductor heteroepitaxy fab-rication and processing community, such as ReflectionHigh-Energy Electron Diffraction (RHEED), to the in-herently surface-sensitive probes of magnetism, such asspin-polarized electron spectroscopies, to traditionalprobes of bulk magnetic materials that are adapted toenhance their surface sensitivity, such as Mossbauerspectroscopy by means of conversion-electron detec-tion. The techniques may make use of major facilitieswith accelerator-based photon sources, neutron scatter-ing facilities, and high-energy electron microscopes.The structural techniques shared with the semiconduc-tor community have been covered in an independentpanel report.117 The present contribution describesmethods used, almost exclusively, by the magnetic-materials community.

While there is a considerable number of sophisti-cated instruments devoted to the analysis of ultrathinmagnetic films, most individual university laboratoriesdo not have the resources to obtain or support all, oreven most of them. In addition, many of these instru-ments do not measure magnetic properties directly, athigh frequencies, or at very rapid rates. And yet manyof the applications, at least for soft or semi-soft mag-netic materials, involve high-speed switching. Simi-larly, scientists need data which are straightforward tointerpret and a rapid turnaround during materialsstudies. The recent developments of the alternating gra-dient118 magnetometer have demonstrated that simple,inexpensive techniques can be developed which havethe sensitivity to extend the range of measurementsto very small samples. Extension of this method toeven smaller samples and/or to high-frequency regimeswould enable many of the smaller laboratories to con-tribute to the understanding of surface magnetism. Inthe same fashion, since surface anisotropy is believedto play such an important role at the interface, in-creased sensitivity improvements in torque magnetome-try, and the ability to use instruments in situ need to bedeveloped. Another excellent example of the usefulnessof a recently developed simple instrument, which canbe used in situ as the sample is prepared, is the surfacemagneto-optic Kerr effect technique.119 Work to de-velop techniques and instruments as simple as theseshould be encouraged.

A. Magnetometry-spectroscopy

1. Mossbauer spectroscopy

Mossbauer spectroscopy can be used as either anin situ or ex situ technique to measure the magnetichyperfine spectra of the magnetic atoms, particularly57Fe. The most useful approach for films is conversion


L. M. Falicov, D.T. Pierce era/.: Surface, interface, and thin-film magnetism

30

20

10

i 20

52oLUa:u-

10

20

10

I I I I I I I I

I I I I I I

I I I I I I I I I

i • i • r

i l l

I I I i i i i

I I I I I2 3 4 5 0 1 2 3 4

MAGNETIC FIELD (kG)

5 6

FIG. 7. Field dependence of magnonfrequencies (crosses) in a representativeset of Ni/Mo superlattices, together withfits to theoretical expressions (solidlines), show the existence of a truesuperlattice effect. The superlatticesare 1-2 /im thick built of bilayers ofmagnetic-nonmagnetic metals withthicknesses of (a) 100 A/300 A;(b) 100 A/100 A; (c) 138 A/46 A;(d) 250 A/750 A; (e) 5000 A/5000 A; and(f) 540 A/180 A, respectively.116

electron spectroscopy, which is carried out in vacuo.One captures the photoemitted electrons and passesthem through an energy analyzer to a detector. Thespectra thus obtained contain not only informationabout the hyperfine field (which is related to the mag-netic moment) but also the quadrupole splitting (indica-tive of deviation from cubic symmetry) and isomershifts (a measure of the conduction electron-spin den-sity at the nuclei). Although Mossbauer spectroscopydoes not directly measure the magnetic moment, thelatter can be indirectly obtained by means of detailedelectronic structure calculations.

2. Magnetic resonance

Magnetic anisotropies can be very precisely andrapidly determined by means of angle-resolved mag-netic resonance. Typically carried out over a range offrequencies from ~10 GHz to 35 GHz and applied mag-netic fields up to 30 kOe, not only the bulk anisotropiesbut in-plane and out-of-plane uniaxial surface an-isotropy energies can be obtained. Furthermore, thetemperature dependence of these anisotropies may be

readily measured. This is important, since generally themagnetic anisotropy energy is much more temperaturedependent than the magnetization itself. Ferromagneticresonance signals have been obtained120 from a sub-monolayer film of Fe grown on Ag(100) and a mono-layer121 of Gd on W. Since what is measured is theintegrated absorbed power within the resonance line,anything which broadens the linewidth ultimately re-duces the precision of the technique. Magnetic reso-nance is therefore also a sensitive measure of thequality of the film as it manifests itself in the magneticproperties. This includes uniformity of the magnetiza-tion, uniformity of thickness, presence of spin-wavescattering sites (cracks, pinholes, imperfections), anduniformity of strains.

3. Magnetometry

There have been isolated efforts to carry out tor-sion magnetometry and ferromagnetic resonancein situ; however, these techniques have not becomewidespread. Magnetometry and resonance are largelyregarded as the primary ex situ characterization tech-niques to obtain important information about the mag-



netization and the anisotropy. A common commercialvibrating sample magnetometer (VSM) has sufficientsensitivity to measure a 20 A, 1 cm2 area Fe film withits full moment of 2.2 /zB- Such instruments typicallyoffer a sensitivity of up to 1(T5 emu. Three orders ofmagnitude in sensitivity can be gained by using a com-mercial Superconducting Quantum-Interference Device(SQUID) susceptometer; however, at such low levelsthe diamagnetic signal from the substrate in generaldominates the data. Very careful procedures must beemployed to remove this diamagnetic signal in order toobtain information which truly represents the film.

More recent developments in magnetometry includea vibrating reed magnetometer122 and an alternating-gradient magnetometer.118 The former takes point-by-point measurements approaching the sensitivity of aSQUID but at lower cost. The latter also approachesthe sensitivity of a SQUID but provides continuousdata similar to a VSM. In addition, a simple torsionmagnetometer, based on a glass fiber, can be readilyconstructed to yield a sensitivity sufficient to measure1 monolayer of Fe on a 1 cm2 surface.

4. Magneto-optics

The magneto-optical Kerr and Faraday effects dateback to the latter half of the nineteenth century. Never-theless, they are now enjoying a renaissance because ofrecent developments in both the basic and applications-oriented communities. On the basic side, it has recentlybeen demonstrated that the Kerr effect can be used todetect monolayer and even submonolayer magnetism.119

The applications-driven opportunities are associatedwith the commercial potential of materials for high-density magneto-optical data storage.123 In addition, therecent development of Kerr microscopy to image mag-netic domains and to observe magnetic-switchingphenomena has helped revitalize the classic field ofmicromagnetics.124125

The Surface Magneto-Optic Kerr Effect (SMOKE)provides a valuable, in situ characterization probe ofthe magnetic and magneto-optic properties of magneticfilms during the growth process. The Kerr effect in-volves the rotation of the polarization of light reflectedfrom a magnetized surface. The magneto-optic cou-pling is caused by the spin-orbit interaction. The tech-nique requires the application of an external magneticfield to reverse the magnetization direction of the sam-ple in the growth chamber. Otherwise, the optical com-ponents are outside the vacuum system. Typically thesystem consists of a laser source, a polarizing analyzer,and a photodiode detector. Magnetic hysteresis curvesare obtained by monitoring the light intensity at thedetector as the field is swept. To address key issuesassociated with the surface magnetic anisotropy, thefield can be in the film plane (longitudinal Kerr effect)

or perpendicular to it (polar Kerr effect). The tempera-ture dependence of the hysteresis loops can be used tomonitor the magnetization and coercivity. Quite re-cently the Kerr effect was used to obtain the magne-tization exponent /3 in the critical regime for the systemFe/Pd(100), and good agreement was found with thatexpected theoretically for a 2-D Ising system.126 TheKerr effect can be used as well to monitor the Curietemperature as a function of thickness, which providesa fundamental characterization parameter of the filmsof interest.

In the future it should be possible to use tunablephoton sources in the optical-frequency region to moni-tor the Kerr rotation of magnetic monolayer and ultra-thin, metastable phases. This form of Kerr spectroscopywill provide electronic structural information in theform of a joint density of states weighted by magneto-optic matrix elements.127 The spectral informationshould complement that obtained from fc-dependentprobes of the band structure, such as angle-resolved,spin-polarized photoemission.

The Kerr effect is not an inherently surface-sensitive probe. The optical penetration depth in metalsis —100-200 A. The surface sensitivity is derived fromthe sample fabrication techniques that create extremelythin epitaxial magnetic films. It is of interest to usecomplementary techniques with different probingdepths to understand coupled magnetic layers, for in-stance. It should be possible to develop the Kerr effectinto such a probe by using nonlinear optical processes;surface sensitivity will be obtained by monitoring theKerr rotation in the Second-Harmonic Generation (SHG)mode.128 The SHG technique has recently gained promi-nence as an advanced surface-analysis technique.129

Brillouin light scattering has also proven valuable toobtain the magnetization, and exchange and anisotropyconstants from magnon spectra. These studies can beperformed in situ on overlayers,130 or as a post-growthcharacterization tool on superlattice and sandwichstructures131'132 in air or in controlled high- or low-temperature environments. The information obtained isquantitative and cross-correlates with FerromagneticResonance (FMR) data.133

B. Polarized electron techniques

I. Spin-polarized photoemission spectroscopy

The most direct information on the ferromag-netic electronic structure at surfaces can be gained byspin-polarized photoemission spectroscopy. Early spin-polarized photoemission studies134 measured the polar-ization of the photo yield as a function of photon energywithout energy analysis. Such measurements still havethe advantage that they can be performed as a functionof applied magnetic field perpendicular to the surface


L.M. Falicov, D.T. Pierce et al.: Surface, interface, and thin-film magnetism

up to magnetic saturation of the sample. The intensityof synchrotron radiation permits energy analysis of theelectrons photoemitted from a material magnetized inthe plane of the surface (such as to minimize stray mag-netic fields).135 A movable spin and energy analyzer al-lows investigation along different directions in /c-space.Thus, utilizing the intensity and tunability of synchro-tron radiation for spin, energy, and angle-resolvedphotoemission, one can obtain a complete mappingof the spin-dependent band structure over the entireBrillouin zone.136 With highly focused photon beamsfrom undulators, it will become possible to combinespin-polarized photoelectron spectroscopy with micros-copy to obtain spin-dependent electronic structure in-formation with high spatial resolution.

With the increasing availability of high-intensityVacuum Ultraviolet (VUV)/soft x-ray radiation basedon insertion devices, spin-polarized photoemissionspectroscopy will play an increasing role in magneticmaterials research. Studies of surface shifts in shallowcore levels, e.g., 4/ levels in rare earths, allow one to dis-tinguish a magnetization at the surface different fromunderlying layers.137 At x-ray Photoemission Spec-troscopy (XPS) energies, the polarization of electronsemitted from multiplet split core levels, such as a 3s or3p level in Fe, gives element-specific magnetic informa-tion.138 In this sense it would be similar to polarizedAuger spectroscopy, but possibly easier to interpret.Furthermore, it may be possible to extract quantitative

values for atomic magnetic moments at surfaces fromthe spin-polarized XPS measurements.139

2. Polarized Auger spectroscopy

The strength of Auger electron spectroscopy as asurface analysis technique derives both from its surfacesensitivity and the fact that Auger electron energies areelement specific. In the case of a ferromagnet, theAuger electrons may also be spin polarized. The spinpolarization results from the different occupation of thespin-split valence-conduction electrons; when theseelectrons at the top of the Fermi distribution are di-rectly involved in the Auger emission process the emit-ted electrons are naturally polarized. If, on the otherhand, only core levels are involved, there may still be aspin polarization because of the exchange interactionof the valence-electron spin density with the filledcore levels. Through spin-polarized Auger, one has anelement-specific probe of the local magnetization at agiven site. Spin-polarized Auger spectroscopy is usefulnot only for investigating the magnetic properties of asurface, but it can also provide information (in films ofa few layers) on the magnetic properties of substratelayers near the interface.

Some features of spin-polarized Auger spec-troscopy are illustrated in the investigation140 of themagnetic coupling of a monolayer of Gd evaporated onan Fe(100) crystal surface. The spin polarization of theFe and Gd Auger lines shown in Fig. 8 have opposite

p -

5 -

0

-5 -

-10 -

-15 "

1

•

-

: J; t

7• ft

•if

-

M

1

Af

Fe23M45M45

, 1 , , i

I

Q

\\4

, 1

Gd

1

N67 N 4 5 N 6 7 N 6 7res

?%

\*

1 1 1

onant

^ _

i, i , , , ,

FIG. 8. Spin polarization versus kineticenergy of secondary electrons, includingthe labeled Auger transitions, from aGd film on Fe(100), excited with pri-mary electrons of 2500 eV. The filmthickness is ~1 monolayer (2.4 A);T = 150 K.140

0 50 100 150 E (eV)



sign, indicating that the magnetic moments in the Gdoverlayer are coupled antiparallel to those in the Fesubstrate. In the same investigation it was possible tomeasure independently the temperature dependenceof the magnetization of the Gd layer and the Fe inter-face layers, taking advantage of the Auger elementalspecificity. For electron kinetic energies below 20 eV,i.e., in the secondary-electron range, the electron polar-ization is seen to be negative. This is surprising sincelow-energy secondary electrons from Fe have a positivespin polarization and is perhaps indicative of the spin-dependent inelastic scattering, anomalously large inthis case, discussed in the section on polarized sec-ondary electron emission.

3. Spin-polarized low-energy electron diffraction(SPLEED)

Low-energy-electron diffraction (LEED) is one ofthe standard techniques to study the structure of sur-faces. Surface reconstructions and relaxations havebeen studied in great detail for clean single-crystal sur-faces, and the geometry of adsorbates has been estab-lished for many systems.141 Using a spin-polarizedelectron beam (SPLEED) on a ferromagnetic surface,one can also gain information on the surface magne-tization through the additional exchange interactionpotential.142 For instance, the temperature dependenceof surface magnetization in the magnon regime hasbeen studied on a surface of a metallic glass,143 and thecritical exponent on Ni single-crystal surfaces has beendetermined.144 More recently, the critical behavior ofthin epitaxial Fe layers has been measured.145 SPLEEDhas also been used to measure the surface magneticmoments on Fe and Ni surfaces.146 SPLEED studies onGd surfaces showed an enhanced surface Curie tem-perature TCs and indicated an antiferromagnetic sur-face coupling.137

Since the strong interplay between structure andmagnetism is well known, it would be highly desirableto combine a structural tool with a probe of the magne-tization. SPLEED contains information on the struc-ture and magnetization simultaneously. QuantitativeLEED structural analysis requires the comparison ofintensity versus energy spectra on a number of dif-fracted beams with the results of multiple scatteringcalculations. Experimentally, large amounts of data canbe accumulated by using Video-LEED systems. Itwould be possible to convert such a system into a VideoSPLEED by adding a spin-polarized (e.g., GaAs) elec-tron gun.147 This would allow the detailed structuralanalysis and magnetic structure determination of anumber of interesting systems. For example, one wouldbe able to study the layer-dependent magnetizations atsingle-crystal surfaces in great detail (also their tem-perature dependence). The structure and magnetic

moments of monoatomic layers can be studied. Also,SPLEED can give information on the magnetic mo-ment distribution in epitaxial ultrathin films, e.g., dis-tinguishing moments at the interface, surface moments,and center-atom moments, putting state-of-the-art elec-tronic structure calculations to a test.

4. Spin-polarized secondary electron emission(SPSEE)

When bombarding a surface with high energy elec-trons (greater than a few hundred eV), a large num-ber of low-energy secondary electrons is emitted. Thislow-energy secondary-electron cascade is produced bymultiple inelastic scattering. If the sample has a netmagnetic moment, the secondary electrons are spin po-larized. It is well established that the direction of thespin polarization is aligned with (and opposite to) thedirection of the magnetization. Therefore, secondaryelectrons can be used to determine the magnetizationdistribution at a surface.

The expected polarization of the low energy "true"secondary electrons, to the extent they are a represen-tative sample of the valence electrons, is estimated tobe P = nB/n, where nB is the magnetic moment peratom (Bohr magneton number) and n is the number ofvalence electrons. The polarization is expected to be28%, 19%, and 5% for Fe, Co, and Ni, respectively.

There are two unexpected features in SPSEE:(1) The spin polarizations of the very low energy elec-trons (<10 eV) are enhanced by a factor of two or threecompared to the average valence band polarization.This has been established for all three ferromagnetic3d-transition metals.148 (2) The surface sensitivity is ap-parently much greater than expected from the "univer-sal" electron mean-free path curve. The magneticprobing depth in Ni and Fe is found149 to be only of theorder of 5 A, which makes SPSEE an attractive tech-nique for the study of ultrathin film systems. It has beensuggested that both effects have their common origin instrongly spin-dependent inelastic scattering.148 There isa need for a complete theory that would allow one tocalculate inelastic scattering, with inclusion of the ex-change interaction.

By using highly focused primary beams (electronmicroscopes), the magnetization at a surface can bemapped with high lateral resolution (100 A) (see thesubsection below on SEMPA). Even without the spatialresolution provided by the Scanning Electron Micro-scope (SEM), the measurement of the polarization ofsecondary electrons provides a strong signal and agood way to obtain information on the average magne-tization at a surface, as opposed to the small samplingof particular transitions over a small region of k spaceas observed in polarized photoemission, and withoutthe complications introduced by diffraction or multiple



scattering as in Spin Polarized Low Energy ElectronDiffraction (SPLEED).

Recently SPSEE has been applied to study thetemperature dependence of the magnetization in thinfilms in the spin-wave regime150 and for thin Fe layersin the critical regime.145 Also, SPSEE gives informa-tion on the reduction of the average magnetizationupon adsorption.149

5. Spin-polarized electron energy-loss spectroscopy(SPEELS)

It has recently become possible to measure spin-dependent electronic excitations in ferromagnets by spin-polarized electron energy loss spectroscopy (SPEELS).SPEELS has been applied to Ni, Fe, and Co sur-faces.151152 When in addition to using a primary polar-ized beam the polarization of the scattered electrons isalso measured, an unambiguous deconvolution of thescattering processes into "flip" and "non-flip" channelsis achieved. It was shown that for Fe and Ni exchangescattering constitutes a significant part of the totalenergy-loss processes. These data also show that theenergy-loss rate in ferromagnets can be very spindependent. In particular, the energy-loss probability forspin-down electrons in Ni can be four times higherthan for spin-up electrons. These findings have certainbearing on the spin dependence of the electron mean-free path in ferromagnets and provide a possible expla-nation for the polarization enhancement and surfacesensitivity in spin-polarized secondary electron spec-troscopy. The probing depth (5 A in 3d metals) mightbe determined by the mean-free path for inelastic ex-change scattering. In other systems (rare-earth metals)the situation is very unclear. Measurements on rare-earth overlayers seem to indicate very strong exchangescattering leading to a probing depth of only ~1 mono-layer. It is obviously very important to understand thespin-dependent scattering mechanisms in these systemsin order to interpret secondary-electron polarizationand polarization in other types of spin-polarized spec-troscopies (e.g., photoemission).

The SPEELS experiment with polarization analysisdoes not require a ferromagnetically aligned sample,since spin-flip transitions can still be detected by achange in the polarization (depolarization). This isequivalent to polarized neutron scattering where, forexample, magnons above Tc can be detected. RecentlySPEELS on Ni above Tc revealed inelastic spin-flipscattering, as shown by a strong depolarization. This isevidence for the existence of local moments in theparamagnetic state and the persistence of a spin-splitelectronic structure151 above 7c-

Since no long-range ferromagnetic order is re-quired, SPEELS can also be applied to other systems,like antiferromagnets. It may be possible to measure

the exchange splitting on a Cr(100) surface, which ispredicted by theory to be on the order of 2 eV becauseof the large enhancement of the magnetic moment (2-2.5 ju.B) at the surface. The energy resolution currentlyachieved in SPEELS can be improved to less than10 meV. This would open up the field to study collectivespin excitations (magnons) at surfaces by SPEELS.

6. Polarized particle probes

In addition to the above spin-polarized adaptationsof conventional electron spectroscopies, there are addi-tional polarized-particle probes of surface magnetism.These utilize spin-polarized atom,153 ion,154 or positron155

beams. The atom and ion beams are sensitive to theoutermost layer of the surface region. They consist ofspin-polarized metastable helium He(23S) and grazingincidence (—150 keV) polarized deuterons, respec-tively, which impinge on a magnetized surface. Thede-excitation of the atom beam involves interatomicAuger processes. The emitted electrons have an asym-metry which depends on the spin orientation of the probeatom with respect to the magnetization of the targetsample. In Electron Capture Spectroscopy (ECS)154 thedeuteron neutralization is detected by a nuclear reac-tion that yields 4He particles whose angular-distributionasymmetry, caused by hyperfine interactions, providesa measure of the spin polarization of the captured elec-trons. In both spectroscopies the detected asymmetriescan be studied as a function of temperature, crystalface, chemisorption, etc. to obtain surface-magnetisminformation. ECS has been used quite extensively to ad-dress many of the major issues in the field; on the otherhand, spin-polarized metastable-atom de-excitationspectroscopy, like polarized positron scattering, hasbeen demonstrated only in feasibility studies.

C. Electron microscopy

Electron microscopy offers the opportunity to char-acterize both the magnetic structure and the atomicstructure of materials in the same optical column. Itsspatial resolution is its most significant advantage, andthe magnetic behavior of a material can be directlyrelated to heterogeneities of both structure and compo-sition on a near-atomic scale. Direct observation ofdomain wall pinning is possible, for example, and theatomic structure of the pinning site can be completelydetermined. There are no new requirements imposedon samples for imaging of their magnetic structure(standard microscope specimens can be used), althougha change in operational mode of the microscope isessential, as described below.

Characterization of the localized magnetic struc-ture of a sample includes the direct determination ofmagnetic domain size and morphology, the structure of


L.M. Falicov, D.T. Pierce era/.: Surface, interface, and thin-film magnetism

domain walls, the location and strength of stray fields,and the magnetization direction of all magnetic fea-tures. It might also include the dynamic observation ofhow the magnetic sample responds to the application ofexternally applied fields, increasing temperature, me-chanical stress, and the like.

Characterization of the localized atomic structureof a sample includes its local crystal structure and ori-entation, grain size and morphology, defect structures(including dislocation, stacking faults, twins, grainboundaries, voids, and inclusions), second-phase par-ticles (including their structure, composition, andinternal defect structure), compositional variations(e.g., segregation at internal interfaces), and the atomicstructure of surfaces and interfaces.

To conduct these studies in the electron micro-scope involves operating the imaging systems in a waythat is sensitive to the localized magnetic fields withinthe sample, and comparing the resulting images to themore traditional images formed by scattered electrons.More recently, the use of a detector that is sensitive tothe spin polarization of the electrons has been utilizedfor imaging domains.156

1. Scanning electron microscopy with polarizationanalysis (SEMPA)

For many purposes, it is desirable to have a high-resolution domain imaging technique in which the con-trast is proportional to the magnetization, as in imagingby the magneto-optic Kerr effect, but not be con-strained by the resolution limitation imposed by thewavelength of light. Further, one wants an imagingtechnique that can be applied to thick specimens inorder to image magnetic structure on a nonmagneticsubstrate, such as a bit written on a magnetic disk ora permalloy memory element on a silicon chip. Thiswould avoid the need for thinning the specimen, as re-quired for Lorentz microscopy on the transmissionelectron microscope (TEM), which is not only tediousbut can change the magnetic properties to be studied.Such a high resolution imaging technique157158 is real-ized in SEMPA. By measuring the spin polarization ofthe same secondary electrons which form the ScanningElectron Microscope (SEM) topographic image, onesimultaneously obtains an image of the magnetizationwith the high resolution of the SEM. All three compo-nents of the magnetization can be measured. Becauseof the inherent inefficiency of currently available spinanalyzers,156159 the polarization measurement takes ap-proximately 104 times as long as an intensity measure-ment of comparable precision. The resolution at presentis 40 nm. In the near future sub 10 nm SEMPA resolu-tion is expected for an SEM with a field-emissioncathode. The secondary electrons sample at most theouter few nanometers of the specimen so that SEMPA

is sensitive to the magnetic microstructure at or nearthe surface. This is an extra advantage for studying sur-face and thin-film magnetism.

An example of a SEMPA image160 of a test patternwritten on a thin film hard disk is shown in Fig. 9. Themagnetic material is a 70 nm thick film of approxi-mately 80% Co-10% Ni. A low magnification image isshown in Fig. 9(a). The light and dark stripes orientedapproximately horizontally comprise the test pattern ofwritten domains or bits. The bits were written succes-sively in tracks, seven of which are seen running verti-cally. Domains of antiparallel magnetization appearas alternating black and white areas. The magnetiza-tion lies in the plane of the film, as indicated by thearrows in the higher magnification image shown inFig. 9(b). The three nearly complete horizontal bandsin Fig. 9(b), two dark and one light, are domains whichat this magnification are seen to have irregularboundaries. Information is associated with the transi-tion from one domain to another. A sharp, well-definedboundary is desirable for the minimum noise signal (seeSect. V, Applications). The jaggedness of the domainboundary clearly puts a limit on the maximum record-ing density. At the same time that one measures thecomponents of the magnetization, one also obtains theconventional topographic image from the secondary-electron intensity, as shown in Fig. 9(c). This intensityimage is for the same area as the magnetization imageof Fig. 9(b) and shows the grooves commonly found ona hard disk. The SEMPA magnetization image is inde-pendent of the topography, which is an advantage rela-tive to Kerr or Lorentz microscopy magnetic imagingwhere topographic and magnetic contrast can be diffi-cult to separate.

2. Lorentz electron microscopy

Lorentz microscopy exploits the Lorentz force ex-erted on the imaging electrons by the internal and strayfields associated with a magnetic sample, and can beapplied in scanning or transmission modes. Recallingthat in the SEM the scattered electrons are collected bya detector that sits above the surface of the sample, theLorentz force affects both the secondary-electron signal(particularly from the stray fields above the sample sur-face) and the backscattered electron signal (particularlyfrom the influence of the internal induction of the back-scattered yield). These two effects result in "Type 1"contrast and "Type 2" contrast, respectively, and havebeen used for some time.161

To obtain contrast from Lorentz-scattered electronsin the transmission electron microscope (TEM), they caneither be blocked with an objective aperture (Foucaultmode), displayed by phase interference through de-focusing the objective lens (Fresnel mode), or exhibitedin holographic fashion, also by phase interference.162



FIG. 9. The SEMPA image in (a) shows approximately horizontallight and dark bands corresponding to a written test pattern ofmagnetic bits, one of which is outlined near the center. In (b) aten-times-higher magnification magnetization image than in(a) shows the irregularity of the domain boundaries which con-tribute to the read-back noise and ultimately limit the density ofinformation that can be recorded. The intensity image in (c) showsthe surface topography of the same region as in (b).160

The last technique shows lines of constant induction inthe sample, but places severe restrictions on the thick-ness (and flatness) of the sample.

In the scanning transmission electron microscope(STEM), the Lorentz-deflected electrons are best de-tected in the far field, so that the signal at the detectionplane is stationary, even though the incident beamis scanned. This can be done by methods similar tothe TEM, or by the use of a split detector that formsan image on the basis of Differential Phase Contrast(DPC).163-164 A display of the difference signal betweensegments of the split detector shows the regions of thesample that provide the Lorentz force on the imagingelectrons. Because such a signal may have any directionwithin the plane of the specimen, it is important to beable to rotate either the sample or the detector. A moreacceptable option is to divide the detector into manysegments (quadrants at least) and scan through the de-tector signals until a difference image is detected. Withelectronic manipulation of the signals, a variety ofmicromagnetic information can be obtained, includingthe quantitative assessment of induction integratedover the electron path.

The most severe limitation on spatial resolution inLorentz microscopy methods is the need for a magneticfield-free region around the specimen. Electron micro-scopes rely on electromagnetic lenses for resolution(probe-forming in SEM, aberration correcting in theTEM), with optimum resolution coming from strongfields in which the specimens are immersed to maintainshort focal-length conditions. Obtaining a field-freecondition requires that the microscope be run in longfocal-length conditions, sometimes achieved by actuallyturning off the objective lens in the TEM or STEM.This unsatisfactory condition is being addressed withthe design of new field-free lens configurations,165 andby the use of field-emission guns for better probes dur-ing beam scanning. Nevertheless, the current resolutionlimit for Lorentz methods in electron microscopy is ap-proximately 3 nm.

3. Conventional electron microscopy

The most attractive aspect of Lorentz microscopyis that it is readily complemented by conventionalmethods of microstructural evaluation, within the sameinstrument. By switching to normal imaging modes inthe SEM, TEM, or STEM, microstructural informationis rapidly obtained, at the superior spatial resolution ofthese methods. In TEM, atomic resolution is nowachievable, and can be applied to image most magneticmaterials, as long as the volume of magnetic material isnot so large as to aberrate the imaging beam. Further-more, complementary diffraction and spectroscopicmethods popular in electron microscopy can also beapplied. These include selected area diffraction, micro-


L. M. Falicov, D. T. Pierce et ai: Surface, interface, and thin-film magnetism

diffraction, convergent beam electron diffraction,energy-dispersive spectroscopy of x-rays, and electronenergy loss spectroscopy. Spatial resolution in thesemethods is limited by spot size, which is currently inthe 2 nm regime. In the SEM, crystallographic informa-tion can be obtained by electron channeling patterns,and morphological information from standard imagingprocedures. Compositional maps that show local spec-troscopic information at the 50 nm range are univer-sally attainable.

It is still necessary in TEM and STEM methods tothin the sample to the condition of electron transpar-ency, often with specific geometrical constraints so thatthe electron beam traverses an interfacial region incross section, for example. Many methods are availablefor such sample preparation, and they can be applied tobulk, thin film, or multilayer magnetic materials withno modification. Even if exact geometrical orientationis missed during sample preparation, the microscopegoniometer can usually be used to adjust the orientationduring imaging. Finally, the use of computer modelingof the image formation process lends credibility to theinterpretation of even the most complex images, andthe current trend is to have such simulations availableon-line.

4. Scanning tunneling microscopy (STM) andmagnetic force microscopy (MFM)

In the last few years the STM has emerged as apowerful means to study surface structure at the atomiclevel. Rather spectacular results have been obtained forsemiconductor- and metal-surface reconstructions andfor adsorbates on such surfaces. So far the STM has notbeen used significantly to characterize growth, al-though it has the potential to answer such an importantquestion as whether a uniform, continuous monolayerof material, a prototype two-dimensional metal film,has been achieved. There is a great opportunity to cor-relate STM results with those of other techniques inefforts to characterize more fully growth of magneticthin films.

A natural question is whether it is possible to ex-tend the STM to include spin sensitivity so that spinconfigurations can be imaged with atomic resolution.The possibility of using a magnetic tunneling tip, suchthat the tunneling electrons are polarized and must findempty states of the same spin to tunnel into, has beenconsidered.166 Several hurdles to achieving this spinsensitivity have also been considered. The magneticelectrons, for example d electrons, are more localizedthan s-p electrons and thought to tunnel about a hun-dred times less efficiently. Furthermore, there are ex-pected to be strong interactions between magnetic tipsand samples, such that the spin configuration to bemeasured could be significantly distorted. Ideally one

wants to control the polarization of the tip electron andreverse it at will. An optically pumped GaAs tip mayovercome some of these problems, but formidable ef-fort will be required to implement the technique.

The MFM uses a fine magnetic tip on a cantileverof small spring constant to detect variations of the mag-netic field or field gradients just above the surface.167

The MFM suffers from the same problem as a magnetictip STM in that there is a perturbing interaction be-tween tip and sample.168 It does not have the high reso-lution gained by tunneling; it is limited by tip size, andlateral resolutions of 1000 A may be expected. TheMFM is in many ways similar to the Bitter technique.It has the advantage that it can operate in air, and itsenses the stray magnetic field which is the informa-tion wanted for some magnetic applications. It doesnot appear to be well suited for studies of domainwall structures or to obtain information on the samplemagnetization.

D. Diffraction

Conventional diffraction techniques have beenused for many years to determine the structure of bulkmaterials. As such, these techniques are well estab-lished and therefore can be used reliably. The applica-tion of diffraction techniques to thin films, surfaces,and interfaces is limited by the small amount of ma-terial available in the sample and the complicationscaused by the presence of substrates. These difficulties,however, can be overcome by the use of more intenseradiation sources, and by a more detailed understand-ing of the structure of the substrate.

1. X-ray diffraction

It is quite clear that the magnetism of surfaces,interfaces, and films is intimately connected with theirphysical structure. Therefore, magnetic studies are ofdoubtful validity in the absence of complementarystructural information. While neutrons, x-rays, andelectron diffraction can play useful roles, x-ray analysisis the best established and probably the most powerfulprobe of overall structural characterization. Well estab-lished techniques exist to deal with structures of greatcomplexity, with defects, and with structural rearrange-ments. Furthermore, new synchrotron x-ray sourcesprovide enormous intensity, thereby opening new av-enues to the x-ray study of chemical, and even mag-netic, structure.

Determination of the structure of interfaces in thinfilm systems is a difficult problem requiring the devel-opment of new methodologies and the refinement ofolder ones. Defects such as roughness, interdiffusion,and dislocations become major determinants of the dif-fracted intensity. It is precisely the same structures that


L.M. Falicov, D.T. Pierce et al.\ Surface, interface, and thin-film magnetism

dominate, in many cases, the magnetic properties. Al-though the problem has received considerable atten-tion,

42,169,170 most studies begin with a model of thedisorder, include it in a structure-factor calculation,and compare the results with experiment. For example,the Fresnel formalism has been used extensively to in-terpret small-angle diffraction data from multilayeredfilms,171172 with roughness parameters introduced phe-nomenologically through a pseudo-Debye-Waller factor.For true superlattices (crystalline films), on the otherhand, standard kinematical theories have been appliedto treat the effect of disorder on large-angle diffractionpeaks.173 While these theories give qualitative predic-tions, a quantitative understanding is lacking. In sys-tems where large- and small-angle data are bothavailable, for example, it has not been possible to pro-duce a model that brings the two sets of data into quan-titative agreement.174

Recently a nonlinear optimization method, similarin spirit to the extensively used Rietveld refinementmethod, has been applied to x-ray diffraction data fromfilms and multilayers.172 In this approach, continuousand discrete roughness, both perpendicular to, and inthe film plane, interdiffusion, and polycrystallinity areincluded as adjustable parameters for refinement. Veryhigh quality fits can be obtained in this fashion fordiffraction data from superlattices. An alternativeapproach uses a diffraction model for multilayers,decreasing higher Fourier components of the composi-tion and lattice parameter variation by adjustable damp-ing factors.

Before these techniques become standard, wellcontrolled experiments should be performed on sam-ples with induced, controlled disorder. The resultsshould be compared with neutron-diffraction andelectron-diffraction results. The latter are particularlyimportant in helping to distinguish local random rough-ness, such as caused by fluctuations in growth condi-tions, from correlated roughness caused by systematicdrifts in preparation conditions.

2. Neutron scattering

Although neutron scattering can give, in principle,the same structural information as x-rays, it suffersfrom lower intensity and resolution and is, of course,tied to major facilities. At the present time, however,neutron scattering is the method of choice for deter-mining magnetic structure and, especially, detectingmagnetic excitations. The magnetic cross section forneutron scattering, while small, still permits determina-tions of thin-film and multilayer magnetic structures.Conventional triple-axis methods, with moderate neu-tron fluxes, have been used successfully to determinethe detailed magnetic structures of Dy/Y and Er/Ymultilayers,175 and of Er films61110 as thin as 30 nm

(—10 6 cm3 of Er). By increasing the area of the filmfrom 1 cm2 to 10 cm2 and working at the highest fluxcurrently available, it should be possible to extend suchmeasurements to the 1-3 nm regime. Substrate back-ground becomes a major factor, requiring energy analy-sis and a reduction of substrate volume to enhance this.Polarization analysis increases sensitivity further byseparating magnetic and nuclear scattering and is par-ticularly important for ferromagnetic films. This hasbeen especially useful in studies of Gd/Y and othermultilayers.62'107

Analysis of magnetic neutron data requires simulta-neous knowledge of the chemical structure. In theDy/Y and Er/Y work, the structure was modeled by aFourier series for a square wave with damping of suc-cessive terms.39 Simultaneous treatment of nuclear andmagnetic peaks with different damping factors for com-position, lattice spacing, and magnetic modulationpermitted a layer-by-layer determination of magneticmoment and orientation in these modulated magneticstructures. Similar procedures can be used to model thestrain distribution in thin films if data are taken at nu-merous reciprocal lattice points.

Techniques are becoming available to probe thedepth dependence of the magnetization in thin mag-netic films or at the surface of bulk magnetic systems byneutron scattering. This technique, Polarized NeutronReflectometry (PNR), was developed by G. P. Felcherat the Argonne National Laboratory and involves re-flecting spin polarized neutrons at grazing incidencefrom the surface of the specimen.176 The reflection ofthe neutron beam can be described by a spin dependent,depth (z) dependent refractive index of the specimen,

/»*(*) = l-c{b± B(z)},

which includes contributions from the nuclear and mag-netic neutron scattering. For typical materials n differsfrom unity by 1 part in 105, which gives critical anglesfor total reflection of less than one degree. The experi-ment consists of measuring the reflectivity of spin upand spin down neutrons at a fixed incident angle as afunction of neutron wavelength for wavelengths up tothose for which the neutrons are totally reflected. Thereflectivity, calculated from models of n±(z), is com-pared to the experimental reflectivity curves.

The first use of PNR in magnetic materials177 wasto determine the magnetization profile of a sputter-deposited film of Fe3O4 approximately 2500 A thickand to compare it to that of the same film after furtheroxidation to y-Fe2O3. Surprisingly, the reflectivitydata on the annealed film showed that there was a non-magnetic layer at the surface about 150 A thick. Laterthe magnetic inhomogeneity of the film was shown toarise from the formation of a nonmagnetic a-Fe2O3

phase at the surface of the film. The distribution of the



nonmagnetic phase through the film, as determinedfrom grazing incidence x-ray diffraction, was shown tobe in very good agreement with the detailed variationof magnetization deduced from the PNR data, provid-ing confirmation of the power of the PNR method.178

The primary merits of the PNR technique are thatit is nondestructive and gives the absolute magnitude ofthe magnetization for depths up to about 5000 A fromthe surface. There is no lateral resolution in PNR, andit requires very flat samples about 1 cm2 in area. How-ever, PNR can probe magnetic layers buried beneathseveral hundred angstroms of nonmagnetic or antiferro-magnetic layers. For example, recently the magneticstructure of permalloy layers comprising part of anexchange coupled structure of the form Si(lll)/NiFe(400 A)/FeMn (400 A)/ Ta(200 A) has been successfullyprobed179 using PNR. It is thus established that thistechnique is able to probe buried magnetic interfaces.At present the depth resolution of PNR is limited toabout 20 A. However, in recent experiments it has beenshown that it is possible to determine the magnetiza-tion of ultra-thin layers of Fe and Co, enhancing thespin asymmetry of the neutron reflectivity by coveringthe magnetic layer with a thin layer of Cu, which actsas an antireflection coating.180 The enhancement of thespin asymmetry of R± depends sensitively on thethickness of the overlayer. With this method the mag-nitude of the magnetic moment of Fe and Co layersonly one monolayer thick was determined.

Since PNR is sensitive only to the component ofmagnetization normal to the scattering plane, by orient-ing the magnetization of the sample perpendicular tothe sample plane the neutron reflectivity becomes spin-independent. It is then possible for the composition ofthe sample to be determined as a function of depthfrom the surface in the same experiment. Such detailedinformation is required in modeling the spin-dependentreflectivity curves to obtain the magnetic structure ofthe film. Finally, the depolarization of the reflectedneutron beam gives information on any lateral inhomo-geneities in the sample, such as those resulting fromformation of magnetic domains of a certain size.

A promising extension of PNR now under develop-ment181 locates a second detector at a fixed angle abovethe reflection plane. By varying the incident neutronwavelength and sample orientation it should be possibleto perform surface neutron diffraction from surfacemagnetic structures. The scattering wave vector in thiscase lies in the plane of the sample, making this probesensitive to moments oriented normal to the sample sur-face. Surface antiferromagnetism and surface magneticreconstruction may be detectable by such techniques.

Because of the very high energy resolution (g 1 /neV)possible, studies of dynamical processes in solids havelong been dominated by neutron scattering. So far,

studies of quasielastic scattering (critical phenomena),spin waves, and magnetostatic modes in thin films havenot been reported, but should be possible with the com-bination of larger sample areas, the use of multiple,identical samples, and increased neutron flux. A seriousproblem here (as in elastic scattering) is the strong scat-tering from the substrate on which the film or multilayeris grown. Triple axis methods can eliminate inelasticscattering from the substrate for structural determina-tion, but only elimination or reduction of the substratescattering will suffice in inelastic studies. Such studieswill complement optical techniques (which are restrictedto small momentum transfer) and electron scattering(which may suffer from multiple-scattering effects).

E. Photon sources

The major new developments in synchrotron-radiation sources open new research horizons in novelmagnetic-film studies. At present the first spin-polarizedphotoemission initiative in the United States has beenestablished at an undulator beamline on the VUV ringat the National Synchrotron Light Source (NSLS) atBrookhaven National Laboratory.136 The superior fluxand brilliance of the undulator source helps to compen-sate for the inefficiency of the spin detector. Thesecharacteristics permit magnetic materials studies toproceed systematically, whereas the earlier Europeanefforts on bending-magnet beamlines were primarilyvaluable to demonstrate the feasibility of the approach.It is expected that spin-polarized band mappings willresult from the synchrotron efforts; the results will testlocal-density functional calculations of the electronicstructure of surfaces and metastable epitaxial phases.

Future developments involve the forthcomingavailability of the Advanced Light Source (ALS) at theLawrence Berkeley Laboratory,182 which will provideadditional undulator-beamline capabilities to satisfy theexpanding needs of the growing community of novelmagnetic-material researchers. A somewhat more spec-ulative advance would involve the availability of free-electron laser (FEL) sources in the VUV/soft-x-rayrange.183 Such sources are being conceptually designedat present.184 The photon-energy tunability is compara-ble to that of synchrotron-radiation (SR) sources, butthe intensity, brilliance, coherence, and monochro-maticity are all projected to surpass substantially theperformance of SR sources. The several orders-of-magnitude increase in intensity would permit magne-tism researchers to perform analogous experiments tothose envisioned by other materials researchers withundulator sources.185 These include pump-and-probeexperiments to study magnetic excited states by syn-chronizing conventional laser sources with the FELpulse train. Also, spin-polarized photoelectron micros-



copy can be envisioned, which would benefit from thesuperior brilliance of the FEL. (The high intensity ofthe FEL source, in addition to providing new opportu-nities, also raises the problematic issue of space-chargeeffects.) The monochromaticity would permit the spinasymmetry at the Fermi energy EF to be obtained forcomparison with transport and susceptibility studies.Only at £ F can the full potential of the FEL-sourceresolution be realized, because at EF there are noAuger processes to introduce lifetime-broadening ef-fects into the spectroscopic results. Still, it is interest-ing to consider whether enhanced resolution wouldenable effects such as magnon sidebands to be observedon core-level spectra.

Core-level spectroscopy will benefit also from theAdvanced Photon Source (APS) at Argonne NationalLaboratory,186 which is projected to be operational inthe late 1990s. Spin-polarized core-level analysis has re-cently been proposed as a means to monitor short-rangedmagnetic order, and is applicable to antiferromagnetsas well as ferromagnets.187 Conventional photoelectrondiffraction from core-level emission also provides anadvanced structural-characterization tool for epitaxialmonolayer-type structures.44 The APS will providemuch needed structural characterization capabilitiesas well, through the use of grazing incidence surface-structure analysis. This is the technique in which x-raycrystallography is performed in the total-external-reflection geometry to enhance surface sensitivity.188

The value of magnetic x-ray scattering to understandthe bulk magnetic structure of the heavy rare-earthspiral spin arrangements has also been demonstrated.189

The technique has also been proved to be effective incharacterizing magnetic superlattices.190 It would be fas-cinating to combine magnetic x-ray scattering with thegrazing-incidence geometry to obtain surface magneticstructures. These studies require the anticipated bril-liance of the APS undulator beamlines, and cannotbe performed at existing hard x-ray sources. Anotherchallenging possibility involves the ability to separatethe spin and orbital contributions to the magneticform factor by means of x-ray scattering.191 It has beendemonstrated, in principle, that such a decomposition ispossible, although experimental confirmation is yet tobe achieved. Since the role of the spin-orbit interactionis so seminal to understanding the surface anisotropyand the magneto-optic response, any additional informa-tion on spin-orbit effects is very welcome. Such studieswould benefit as well from the availability of circularlypolarized x-rays, because of their enhanced magnetic-scattering cross sections relative to linearly polarizedx-rays that are more commonly produced at synchro-tron sources. In summary, the combined approach ofusing advanced synchrotron sources to obtain struc-tural, magnetic, and electronic-properties information

provides extraordinary research opportunities for futureresearch in the field of novel magnetic materials.

V. APPLICATIONS

Many of the applications of magnetic-material sys-tems require control of the extrinsic properties suchas coercivity, orientation, permeability, and micro-magnetic features. Intrinsic properties such as moment,anisotropy, or magnetostriction are normally acceptedas given for a particular material composition. In the lastfew years it has become clear that a better understand-ing of the interactions at the interfaces of materials canbe used not only to control the extrinsic properties butto manipulate the intrinsic properties as well.

While not all of the preparation and characteriza-tion procedures described earlier are the methods ofchoice in a manufacturing environment, the fundamen-tal understanding of the role of the interface developedby the use of these techniques should provide a guidein selecting materials and process conditions for com-mercial applications.

The following are only a few examples of applica-tions of magnetic materials and systems. In each ofthese the ability to improve performance significantlyrequires an understanding of the interface interactionsso that new material systems and practical processescan be developed. It is worth noting that many of thefundamental concepts are common to many of the ap-plications. For instance, the magnetic anisotropy andthe exchange interaction across the interface directlyinfluence the coercivity of both hard and soft magneticmaterials.

A. Magnetic recording

Currently there are many different recording me-dia for recording systems. While videotape systemsoperate at similar track widths to high-performancehard-disk systems—typically 1000-1500 tracks per inch(tpi)—the linear densities are greater by factors of twoto four—80 kiloflux changes per inch (kfci) versus30 kfci. This difference can be justified by the errorrates required and data encoding schemes used. If theperformance-improvement-versus-time curves thathave characterized the last thirty years are to continue,the area densities will increase by a factor of twentyby the turn of the century. This will require not onlythat the mechanical interface between the magnetichead and the media be improved but also that magneticrecording properties of the media be much improved.For example, a longitudinal recording system withtrackwidths of less than 2 pm (10000 tpi) and linearbit densities of 100 kfci will require high coercivities(i=2000 Oe) and very smooth media so that extremelyclose head-medium spacings (=500 A) may be ob-



tained. At these small sampling volumes of a bit,medium noise, head noise, and head output becomemajor magnetic concerns.

1. Hard disk media

Current laboratory media for hard disk thin filmsare composed of a substrate upon which a series of lay-ers of materials are deposited. Typically a nonmagneticunderlayer, such as Cr, is sputtered by means of a dcmagnetron over an amorphous Ni-P layer which hasbeen electroless-plated on an aluminum disk. A numberof Co-based alloys, such as CoNiCr, CoP, CoCrTa, orCoCrPt have been used as the sputtered magnetic layerover the underlayer. Over this layer a nonmagnetic,thin overcoat is deposited to protect the media duringthe time the head comes in contact with the surface.The dominant philosophy is to achieve in-plane an-isotropy, high coercivity, and low noise. The efforts arethus centered around control of interfaces. The non-magnetic underlayer, Cr, is deposited with selectedthicknesses; film growth conditions are chosen so as toprovide grain size control and particular crystal planesare exposed for the magnetic layer to grow on epitax-ially during sputtering. This underlayer provides an in-terface by means of which the hexagonal-close-packedcobalt alloy can be oriented in the plane. The an-isotropy of the alloy provides a coercivity mechanism ifthe grain size is appropriate for a single domain pergrain. The underlayer grain size can be replicated bythe magnetic layer. The magnetic alloy and depositionconditions are such that during deposition various phasesegregations192'193 at the individual grain-grain interfacescan occur. If these interfaces sufficiently decouple themagnetostatic and quantum-mechanical exchange inter-actions between grains, then the roughness of therecorded bit edge will essentially be defined by the ge-ometry of the grains. The medium noise will then bedetermined by grain-counting statistics, just as in goodparticulate media, and not by arbitrary zigzag-shapeddomain wall boundaries.194'195

The very small future bit volumes will requiremuch smaller grain sizes (currently about 1000 A) so asto minimize medium noise. In order to maintain signaloutput, the volume of the nonmagnetic decouplingintergranular boundaries should be reduced so that thetotal magnetic moment per bit is as high as possible.Fundamental understanding of interface decouplinglayers would be helpful, in this context, to guide mate-rial selection. To achieve high coercivity within thesevery small particles, a better understanding of the sur-face anisotropies that can be induced by the surface ofthe nonmagnetic underlayer may be essential. In evenmore futuristic media, as the grain size approaches thesuperparamagnetic limit, surface anisotropy may beneeded to provide the required domain stability.

2. Magneto-optic media

After many years of development, magneto-opticdrives have recently reached the market place. Thesedrives use a medium composed of an amorphous rare-earth transition-metal alloy (e.g., Tb-Fe-Co). The in-formation is recorded by localized heating with a laserand switching the magnetic state. Readout is accom-plished by a combination of the polar-Kerr and theFaraday effects. Recording density performance of thesystem is largely determined by the spot size of the laser(A —800 nm). Bit cell sizes are 1.5 im to 2 jum on aside. The signal-to-noise ratio is largely determinedby the size of the magneto-optic effect and the opticaldepolarization caused by medium imperfections. Thecarrier-to-noise ratio (30 kHz bandwidth) is typicallybetter than 50 dB.

Since the recording densities of current systems arerestricted by the diffraction limit of the light, improve-ment in future systems will require materials whichhave large magneto-optical effects at shorter wave-lengths (A <500 nm). The key issues are to maintainthe perpendicular anisotropy, control the Curie tem-perature, have a reasonably high magneto-optic effect,and obtain a noiseless, nondepolarizing grain structure.Over the years a considerable number of materials withlarger magneto-optic effects than the amorphous rare-earth-transition-metal alloys have been investigated.Almost all, however, suffer from depolarization noise.

Recently, effort to make new magneto-optic sys-tems with compositionally modulated films has shownsome promise.196 These films should have very littlegrain noise. A few angstroms of cobalt layered with sev-eral angstroms of palladium or platinum have beenshown to possess perpendicular anisotropy; thickerfilms do not. The optical effects at short wavelengthsshow promise, but need improvement. Increasing theratio of Co to Pd or Pt would help the optical propertiesbut degrade the anisotropy. In order to make films ofthis type useful for future magneto-optic recording sys-tems, an improvement of the interface-induced an-isotropy and a better understanding of the attendantmechanisms are needed.

3. Exchange biasing of magnetoresistive heads

Recent years have seen increasing use of the mag-netoresistive (MR) read head as a replacement for thetraditional inductive head in the reproduction of mag-netic recordings. Advantages include a velocity indepen-dent signal and lower head noise in some applications.The MR head uses the magnetoresistive effect to changethe voltage across a thin-film element placed near therecording medium and its accompanying fringe fields.This element, typically FeNi, is usually biased by anexternal field so that it operates in a linear regime andconsists of a single magnetic domain. The external field


L. M. Falicov, D.T. Pierce et a/.: Surface, interface, and thin-film magnetism

may be provided by a neighboring permanent magnet;however, exchange biasing through an antiferromagnet(typically FeMn or TbCo) is receiving increasing atten-tion.92197 The major discrepancy between experimentalexchange biasing and simple theory limits the adjust-ment of the biasing for engineering purposes. Thus, aquantitative theory of exchange biasing would be highlybeneficial for this application.

4. High magnetization materials for recording heads

As the recording medium coercivity is raised to in-crease recording densities, it becomes necessary to pro-duce larger fields with the recording head. Saturationeffects, unfortunately, limit the field which may beobtained from a head. Hence, new magnetic thin filmmaterials with higher saturation flux densities thanPermalloy (Bs = 10000 G) are needed. In addition tooffering high saturation flux density, these materialsmust also have zero magnetostriction and be resistantto annealing during processing. There has been consid-erable work on amorphous metal-metal systems (e.g.,Co-Zr) and amorphous metal-metalloid systems (e.g.,Fe-Co-Si and Fe-Co-B). These materials offer satura-tion flux densities of up to 14000 G in nonmagneto-strictive compositions. There has also been work oniron nitride materials, with theoretical saturation fluxdensities as high as 25 000 G, although nonzero magne-tostriction and stability of the materials during anneal-ing are serious limitations. There have been attempts todevelop multilayers with high saturation flux density.Multilayers of Co-Zr/Fe, in which the Fe is kept amor-phous by depositing in extremely thin layers, have beenproduced with saturation flux densities of 18000 G.Again, nonzero magnetostriction, which changes withannealing, is a serious problem. Another future possi-bility is that a suitable growth technique could be foundeither to fabricate superlattices which take advantageof the enhanced magnetism at interfaces or to pro-duce controlled strain effects which increase the mag-netization. An additional possibility could be to employsingle-crystal materials, grown by MBE or simpler tech-niques, which could have a reduced coercivity causedby the absence of pinning states for the domain walls.

B. Magnetoelastic devices

Applications of magnetoelastic phenomena in thin-film devices may exploit two classes of magnetic ma-terials. The first are amorphous transition metals,with nearly zero anisotropy, which exhibit the highestmagnetomechanical coupling factors ever observed;thus they can be used as ultra-sensitive magnetic-fielddetectors with a displacement readout, or strain detec-tors (accelerometers, etc.) with an inductive readout.

These materials can be sputtered in thin-film form. Ahybrid piezoelectric/magnetoelastic structure can beenvisioned which would convert voltage to magneticfield and vice versa.

The second class of materials are rare-earth transi-tion-metal alloys, which have more moderate magneto-mechanical couplings but produce large strains/highpower at reasonable magnetic fields. Such materialsmight be used in thin-film form or in a superlatticecomposite to control the state of strain in a nonmag-netic material (e.g., semiconductor).

C. Integrated optical and electronic devices

Although there has been considerable activity inthe growth of thin magnetic films, there has been littleeffort to introduce magnetic elements into either inte-grated optical or integrated electronic circuitry. Thereare, nevertheless, many opportunities for such elementsto provide nonreciprocal devices to act as isolators,phase shifters, delay lines, circulators, or filters. All ofthese devices are used in current high-frequency tech-nology (microwave signal transmission, radar, etc.). Asthe need develops for ever higher frequency operation,the shorter wavelengths require that the dimensionsof the devices shrink. These devices become so smallthat they must be monolithically incorporated into theintegrated circuits on a microchip. Thin magnetic filmsare, therefore, the appropriate morphology for this newtechnology. One can either use magnetic insulators toact as a dielectric medium within strip-line devices, ormagnetic metals to act as either guidelines or groundplanes for the devices. In either case the challenge isto provide magnetic films on appropriate substrates(Si, GaAs,etc.) having the appropriate magnetizationand anisotropies.198

A second use for magnetic materials is to provide amagnetic field on an integrated circuit substrate. Thereare many devices which require an applied field, but itis impractical to attempt to house a microchip actuallywithin the field of a coil, electromagnet, or permanentmagnet. What is required is the means of providing,on a very small scale (~1 ^tm), a highly localized mag-netic field which affects only a single circuit element.Furthermore, this magnetic field must be providedby a magnetic material, which can be laid down in filmform and readily patterned by techniques which arecompatible with other fabrication techniques commonto the semiconductor microchip industry.

Surfaces and interfaces play an important role inthese applications in determining both magnetizationand anisotropy of the films. For example, as mentionedabove, chemical reactions can diminish the magnetiza-tion at the interface. Appropriate buffer layers at theinterface can prevent these reactions. The nature of the



growth at the interface can often introduce orientedstrains or dislocations which lead to anisotropies. Theseanisotropies may be useful in providing easy axes forthe magnetization or could prove troublesome if onerequires isotropic behavior in the film. Examples suchas these illustrate the importance that details of thefilm growth can have on the ultimate application.

D. Permanent magnets

It has been estimated that by the mid-1990s thecommercial market for the rare-earth-transition-metalalloys as materials for permanent magnets will be sev-eral billion dollars annually.199 The discovery and rapiddevelopment of the Nd2Fei4B class of materials is mak-ing this possible. The emphasis on applications of therelatively inexpensive Nd2Fei4B has provided increasedawareness of SmCo5 and Sm2Co17 materials and is in-creasing the demand for them. Special batches ofNd2Fei4B with energy products approaching 50 MGOehave been made in the laboratory, while materials withvalues approaching 30 MGOe are becoming commer-cially common. Traditional ceramic ferrite magnetshave energy products of approximately 4 MGOe. Theextremely high energy product of the Nd2Fei4B mate-rials has allowed the size and weight of devices to besignificantly reduced.199 For instance, a 100 hp motorwhich normally weighs 1000 lbs can now be reduced insize to weigh about 35 lbs by using these magnets. Simi-larly, the electrical efficiency of a small fractional-horsepower motor can be doubled from 35% to 70%.The high coercivities are allowing significant improve-ments in the package design of planar or pancake-shaped motors and actuators.

These materials are manufactured either by a sin-tered, powder-metallurgy process or by a new rapid-quenching, hot-pressing, and die-upsetting process.Both processes result in very good products. A compre-hensive review article on the rare-earth-transition-metal magnets is available.199 In it, much of the currentunderstanding of the materials and some of the uses forthese magnets are described.

The principal issues for obtaining high-energyproducts are very similar to those described for thehard-disk thin-film media. The magnets need to beoriented in order to utilize their high magnetization.They need to have isolated single-domain grains alongwith a very high anisotropy in order to have coercivi-ties higher than their 4vMs values. The interfacialboundaries and the coupling between grains determinewhether or not domain walls can be nucleated at, orpropagated across, the interface. It is believed that non-magnetic phases must be formed at these interfaces.However, the exact composition and role of these inter-faces and phases are not clear. The temperature depen-dence of the coercivity is very important for many of

the applications, and the role of temperature on theinterface is an open issue.

VI. ISSUES AND PROSPECTS

This section is an attempt to highlight what theauthors consider are important issues, and what theybelieve are the prospects for future research. Openand unsolved problems, current investigations and fu-ture prospects are mentioned throughout this report;some of them are reiterated here briefly. The goal is tobring together, in one section, a brief summary of re-search opportunities in surface, interface, and thin-filmmagnetism.

A. Theory

Ab initio techniques based on the local spin-densityapproximation have been very successful in predict-ing trends in magnetic properties; they need, however,to be applied—as resources and computer capabilitiespermit — to more magnetic-surface and thin-film systems.Details of surface magnetic properties which depend onthe lowered symmetry require techniques which explic-itly represent the solid-vacuum interface by means ofboundary conditions on the electron wave functionsor density—the so-called film codes. These calculationsare extremely time-consuming and expensive, and thecodes are in use at only very few institutions.

Magnetic properties of interfaces between two dif-ferent materials, which also involve lowered symmetry,can be simulated either by film codes or by layeredsupercell calculations which use bulk codes. In eithertype of calculation the spatial scale of properties whichcan be investigated is limited by the number of atomiclayers which can be included in a unit cell, currentlyof the order of 10. For example, investigation of thevery interesting coupling observed between Fe layersin Fe/Cr/Fe sandwiches and Fe/Cr superlattices,which occurs for Cr thicknesses of 10-20 A, will re-quire a very large expenditure of state-of-the-art super-computer time.

Structural relaxation at surfaces and interfacesmakes the limitations of supercomputer resources evenmore acute, as the existence of such relaxations requiressignificantly larger unit cells. In addition, if structuralrelaxations are not known, calculations of many differ-ent structural configurations may be required in orderto determine the minimum-energy relaxation. Thegreatest opportunities in this area of ab initio theory liein the development of more efficient codes for film andlayer calculations, in the development of algorithms forefficient searching of the phase space of structural re-laxations, and in the careful choice of prototype mag-netic systems for study.



The total-energy capabilities of ab initio calcula-tions can be used in a different way to predict magneticproperties of thin films which are caused not by thelowered symmetry but rather by strain or lattice distor-tions due to interface bonding. Bulk codes have provedto be very successful in predicting the systematics ofmagnetic structure—nonmagnetic versus ferromagneticversus antiferromagnetic arrangements—as a functionof lattice parameters and lattice symmetry. Calcula-tions have been performed for most of the 3d magnetictransition metals in the bcc and fee structures. How-ever, hexagonal and lower-symmetry structures—e.g.,tetrahedrally, orthorhombically, or trigonally distortedbcc and fee systems—are just beginning to be investi-gated. These types of calculation, which are relevantfor epitaxially strained or "pseudomorphic" structures,are most useful for films sufficiently thin so that theyremain pseudomorphic but thick enough so that thestrain dominates the surface/interface effects. For thesecalculations to be extended to lower-symmetry systems,the approximations of spherical averaging of potentialsor of electron densities used in the most efficient codesmust be carefully evaluated, and perhaps eliminated infavor of full-potential codes that include nonsphericallysymmetric terms. Elimination of spherical approxima-tions may also be crucial for achieving numerical accu-racy in surface and interface calculations of the typediscussed above.

Theoretical studies should also be pursued to deter-mine the inherent limitations of local-density methodsdiscussed in Sec. II (Theoretical Background). Theselimitations (the few percent errors in the lattice con-stants, for example) seem to be greater in spin-polarizedmagnetic systems. There is preliminary evidence thatmore complicated forms of the exchange-correlationpotential, such as those that include terms in the gradi-ent of the density,200 may improve results in some cases.

In addition to the ab initio calculations, which areinvolved, expensive, and require state-of-the-art super-computers, there is an obvious need to develop simplercorrelations and empirical rules which could eitherprovide qualitative explanations for existing experimen-tal data, or point toward systems and configurationswhich might exhibit some required magnetic property.Attempts in this direction exist,14'77 but they are still toocrude to be of practical significance.

Beyond the calculations of equilibrium structuresand primary magnetic properties (magnetic moment,hyperfine field, exchange splitting, etc.), there is a cru-cial need to determine secondary magnetic properties,such as anisotropy and magnetostriction, by means ofelectronic-structure techniques. These problems arediscussed separately below.

Finally, the richness in structure and the complex-ity of the systems discussed here are, continuously, a

source of surprises for new, unexpected, unexplained,or misunderstood effects which require both qualitativeand quantitative explanation. Theory can develop onlyby the simultaneous paths of constant interaction be-tween theory and experiment, and by the formulationof (by necessity) simple models able to extract, from thelarge number of secondary and irrelevant effects, thebasic features of the phenomenon under consideration.

B. Magnetic moments at surfaces and interfacesThe values of the moments at the surface of mag-

netic metals remain a lively issue which needs more care-ful experimental data. All theoretical calculations76"70

agree with the fact that at free surfaces the magneticmoments tend to be enhanced (in weakly magneticmetals) or created (in almost magnetic metals), eventhough the precise values of those moments tend tovary appreciably from calculation to calculation. Theavailable experimental data, even though not 100% inagreement with each other, tend to confirm indirectlythese theoretical predictions. Experimental confirma-tion, with direct experimental measurement of thespecific surface and/or interface moments, is not yetavailable.

In particular, several issues require further clarifi-cation and reliable experimental data:

(1) The repeatedly calculated and indirectly ob-served large magnetic moment at the {001} surfaces ofantiferromagnetic chromium—a ferromagnetic layer inan ideal, defect-free surface—remains yet to be ob-served directly.

(2) The optically observed (SMOKE) dead layers ofiron [201] when deposited on ruthenium (0001) remainsa puzzling effect which requires careful theoretical andexperimental work. An accurate self-consistent calcula-tion, including structural rearrangement effects, andthe performance of the experiment at low temperaturesis needed.

(3) The magnetic moments, if any, of the free sur-face of vanadium and of vanadium overlayers on a vari-ety of substrates remain an open question.

These are only a few examples of systems whichremain to be examined; the general area is still onlysketchily explored and is a rich ground for basic re-search with possibly many practical applications.

C. Magnetic coupling at interfacesOne of the most exciting areas of both current and

future research is that of coupled magnetic multilayeredsystems. In the simplest case, materials engineering ofthese magnetic systems allows for the optimization andcontrol of such basic magnetic properties as satura-tion magnetization, anisotropy, coercivity, and mag-


L. M. Falicov, D.T. Pierce era/.: Surface, interface, and thin-film magnetism

netic domain structure. It seems clear that these typesof structure will be of increasing importance in themagnetic-recording industry, as finer tuning of thesemagnetic parameters becomes necessary. At the oppo-site end of the spectrum, distinctly new properties ofcoupled magnetic systems have recently been discov-ered in a number of different magnetic systems. Themost recent discovery is that of antiferromagneticcoupling of neighboring Fe layers in Fe/Cr/Fe sand-wiches, together with an enhanced magnetoresistancein such systems.

Examples of areas of research likely to produce im-portant new results include:

(1) A search for new ferromagnetic/metal/ferro-magnetic layered structures displaying antiferromag-netic coupling of ferromagnetic layers. This search willlikely lead to improved understanding of the phe-nomenon, and its related effects in magnetotransportproperties.

(2) Study of the magnetic tunneling valve effect inferromagnetic/insulating/ferromagnetic layered sys-tems, and its dependence on spin polarization of theferromagnetic layers.

(3) Nonequilibrium spin injection in ferromagnetic/metal/ferromagnetic structures, and in particular themagnitude of the spin polarization propagated acrossthe ferromagnetic/metal interface.

(4) Study of ultra-thin antiferromagnetic layers viaexchange coupling to a ferromagnetic probe layer, andin particular the dependence of the magnitude of thecoupling on the atomic-scale structure of the ferromag-netic/antiferromagnetic interface.

(5) Control of magnetic anisotropy in ferromag-netic layers via exchange coupling to a second ferro-magnetic or ferrimagnetic layer.

(6) Study of the magnitude of the exchange cou-pling in ferromagnetically coupled ferromagnetic layersin ferromagnetic/metal/ferromagnetic structures, by de-termination of the temperature dependence of the mag-netization of the ferromagnetic layers.

(7) Fermi-surface driven effects on the magneticand transport properties of magnetic superlattice struc-tures resulting from the imposed superperiodicity.

Exchange coupled magnetic multi-layered struc-tures form a rich area of research. Progress will mostlikely be led by experimentation with different materialcombinations and by attempting to control the micro-structure of the interfacial region between the variouslayers. The ability both to vary the nature of the inter-face in a controlled manner and to characterize the na-ture of the interface provides an extremely challenging,perhaps intractable problem. It might well be that thestudy of extremely small-scale structures, either mag-netic dots or 1-D magnetic chains, perhaps grown onterraced substrates, may provide more homogeneous

structures with which to examine some of the effectsmentioned above.

D. Low-dimensional magnetism

Three prominent issues in surface magnetism con-cern:

(i) the criteria for and impediments to achievingmonolayer magnetism,

(ii) the nature and origin of the surface magneticanisotropy, and

(iii) the critical behavior of 2-D magnetic phasetransitions.

The issues involve the competing influence of elec-tronic and geometric structural considerations. The roleof strain fields at surfaces and interfaces in stabilizingperpendicular easy axes of magnetization, relative tothe role of the spin-orbit interaction, needs to be as-sessed. These studies will benefit from high-qualitysample preparation and the availability of in situ, aswell as post-growth, characterization techniques. Theimportance of growth-induced anisotropies needs to bebetter appreciated. Test cases of well-characterizedmodel systems need to be established.

The area of critical phenomena in low dimen-sions provides a particularly satisfying arena for cross-pollination of ideas between experimentalist andtheorist. This is because the concept of universalityputs the emphasis on characteristic length scales, andnot on the details of the interactions or of the structure.The ability of experimentalists to generate data whichcan be compared to Onsager's 1944 solution of the2-D Ising model is a long-awaited development thatshould be close to realization. Issues associated withfinite-size effects, inhomogeneities, defects, field-induced-fluctuation phenomena above critical tempera-tures, etc. all need to be systematically explored in orderto make meaningful progress in identifying universalbehavior.

Experimental issues associated with the relation-ship between the measurement probe and the magne-tization need elucidation. Invariably surface-sensitiveprobes of the magnetization couple through ill-definedinteraction matrix elements, or are subject to dynamical-scattering effects that introduce intractable correctionswhose influence is difficult to assess. The magnetiza-tion axis is also subject to reorientation as temperatureis raised toward the Curie temperature, especially forvertical easy-axis alignment. The critical region alsoneeds to be defined based on a Ginzburg-Landau crite-rion in order to determine the temperature range overwhich data fitting should take place.

E. Excitations

Thermal excitation of spin waves at surfaces havebeen studied by polarized-electron scattering143 and sec-



ondary-electron emission.202 Experiments confirmedtheoretical predictions203'204 that the temperature de-pendence of the surface magnetization should followthe same T312 power law as in the bulk. The deviationof the prefactor of the Tm term from earlier predictionshas been explained by assuming an exchange couplingstrength of the surface layer to the bulk to be only 30%of the coupling strength between bulk layers.100 Thistemperature dependence of the local magnetizationat surfaces and interfaces caused by the excitation ofspin waves at low temperatures should be further ex-plored as a probe of local exchange interactions in fu-ture studies.

Considerable theoretical effort has been directedtoward the study of magnetostatic coupling of surfacespin waves in multilayer samples. A rich spectrum ofcoupled excitations, the nature of which depends onwhether the ferromagnetic films are aligned parallel orantiparallel to each other, has been predicted.115'205 Ob-servations of additional modes in Ni/Mo films usingBrillouin light scattering supports this picture,116'206 asalso does recent work on Fe/Cr structure.132 The entiresubject of spin excitations in coupled systems is, how-ever, largely unexplored.

Another possible area which has not received suffi-cient attention is a comparison of results from differenttechniques. For instance, no comparison of magnon-mode frequencies extracted from FMR and from Bril-louin scattering exists.

In multilayer systems the induced periodicity foldsthe phonon bands into very small Brillouin zones, witha resulting very complex spectrum. Similar effectsshould be observed in magnetic systems; the resultingmultiplicity of superzone gaps must depend sensitivelyon interlayer exchange coupling. Experimental obser-vation of multilayer magnon bands should be accessibleto neutron scattering techniques, especially as more in-tense sources and larger samples become available.

True surface spin-wave modes (Damon-Eshbachmodes) have been observed in Ni/Mo multilayers andon Fe-Cr-Fe sandwiches by light scattering.116132 How-ever, light scattering can measure only the k = 0 uni-form mode and cannot explore dispersion, which wouldprovide information about surface exchange inter-actions. HREELS may be capable of observing thefull surface-magnon dispersion curve, but requires in-creased energy resolution over that currently available,and better understanding of the scattering processes.

It has been suggested that surfaces order indepen-dently from the bulk. If so, it should be possible to ex-plore both the growth of surface coherence and itsexpansion into the bulk. Quasielastic neutron scatter-ing, perhaps using surface diffraction techniques, is apromising way to pursue such studies. Because truly2-D systems develop long-range order only in the

presence of anisotropy, studies of kinetics of surfacemagnetism provide information about local surface an-isotropy. Kerr effect, magnetometer methods, and spin-polarized scattering may be capable of detecting thekinetics of the ordering process, possibly revealingspin-glass-like surface states or the effects of randomlocal anisotropy.

F. Magnetism and structure

The physical properties of thin films are stronglyaffected by their structure and composition, to the ex-tent that physical properties can often be predicted ifstructure and composition are known with sufficientprecision. It is not always true that structural and com-positional "perfection" are the most desirable traits formagnetic films. Interfacial roughness may, for example,enhance magnetic coupling by breaking wave vectorconservation conditions that would otherwise isolatethe electronic states of two metals in contact.

In general, it is important to appreciate the spatialscale over which magnetic phenomena occur, and toprobe the structural and compositional variations withinmagnetic materials at the same spatial resolution. Di-polar interactions, domain morphology, and domainwalls are "large scale" relative to RKKY and exchangeinteractions, which are "small scale" and relate to struc-ture at atomic dimensions. Each of these requires atten-tion to structural determination at the appropriate scale.

A variety of techniques for structural analysis isavailable which permits the determination of atomic-scale structure using diffraction and macroscopic mor-phology using electron microscopy. For the study ofanisotropy phenomena at interfaces and surfaces,roughness may have an important role. Anisotropy, ofcourse, is crucial for determining many physical proper-ties and for stabilizing magnetic ordering in lower-dimensional systems. Defects may provide sources ofrandom anisotropy and exchange that produce spin-glass-like regions. Coupling across nonmagnetic layersalso relies on structural aspects of the intervening lay-ers. The presence of pin holes, interdiffusion, androughness may modify the details of the interactions;but the detection of pin holes and the distinction be-tween roughness and interdiffusion are very difficult todo experimentally. (The reason is that both roughnessand interdiffusion appear in a similar fashion in thefitting procedures employed to determine the structure;see Sec. IV, Techniques and Facilities, Diffraction.)The preparation of samples with artificially induced,controlled defects is needed to understand the effectthese have on both the structural probes and the physicalproperties. Examples of these types of studies could bethe growth of multilayers on artificially roughened sur-faces, studied by x-ray diffraction, TEM, magnetization,and light scattering, with the objective of understanding


L.M. Falicov, D.T. Pierce et at. Surface, interface, and thin-film magnetism

whether these defects heal or are enhanced as a func-tion of thickness. These artificially roughened surfacescould be produced by growing samples at different sub-strate temperatures, depositing small particles on thesubstrate, using vicinal surfaces, or other procedures.

Generally all types of measurements which corre-late atomic and microstructure with magnetic proper-ties are of considerable interest. Microstructural effectssuch as surface relaxation, surface reconstruction,"roughness", strain (as might be induced by magneto-striction) in thin films, the type and location of defects(including misfit dislocations, threading dislocations,growth ledges, stacking faults), and compositional het-erogeneities (segregation, precipitation, impurities) canall have significant influence on the magnetic behaviorof thin films, overlayers, and interfaces. These must becarefully monitored with high spatial resolution inorder to understand their individual and synergetic ef-fects on local magnetic behavior. Ultimately, the physi-cal structure plays a dominant, perhaps determiningrole in shaping the magnetic properties.

The mechanical properties of thin films and super-lattices have been found to exhibit anomalous behavior,for example the supermodulus effect which is associ-ated with large strains both perpendicular and parallelto the layers. The effect of such mechanical propertieson film magnetism has not received much attention.

Strong magnetoelastic interactions provide an im-portant source of coupling among structure, mechani-cal properties, and magnetism. The strains associatedwith the layered growth of dissimilar materials in con-tact with each other restrict the ability of magnetic ma-terials to distort in response to the magnetoelasticenergy. Systematic studies are needed to determine thedefect structure of such films and multilayers and thechanges that occur upon magnetization. The differ-ences that accompany compression and expansion ofmagnetic films caused by differing lattice parameters isa particularly fruitful direction for future research.

G. Metastability

The work on metastable magnetic structures de-scribed in Sec. Ill (Materials) actually represents thebeginning of what should prove to be a fruitful area ofresearch. The initial work on bcc Co and fee Fe focusedonly on cubic phases. More recent work has now madeit clear that cubic phases are a special case, and that itis more generally expected that body-centered tetrago-nal phases will be the metastable phases most likely tobe stabilized by epitaxial growth. This is true both froma theoretical point of view—where recent calcula-tions207 found a body-centered-tetragonal metastablephase for Cu—and from an experimental perspective.

Experimentally, the number of possible single-crystalsubstrate materials is limited. Except for the relatively

rare cases of a lattice matched to a cubic phase, theactual growth should in general compensate for thein-plane mismatch by relaxing the interplanar spacingduring growth. The resulting tetragonally distortedstructure should achieve stability if it is energeticallyclose to a metastable tetragonal phase. The challenge,from the experimental viewpoint, is to characterizecarefully the structure of these new tetragonal phasesby LEED, EXAFS, or other suitable techniques, andto measure their magnetic properties as well.

H. Anisotropy and magnetostriction

An understanding of anisotropy and magnetostric-tion in transition-metal materials is of fundamental im-portance for eventual control and exploitation of theseproperties in applications. Calculation of these spin-orbit related properties should be attacked both byincluding spin-orbit coupling self-consistently in elec-tronic structure calculations and by perturbation-theoryapproaches. The lowered symmetry of surfaces andthin films means that the dominant terms will generallybe second order in the spin-orbit coupling, and thuslarger than in most bulk materials. The important issuesare whether the Fermi surface can be calculated withsufficient accuracy to yield meaningful results andwhether the precision of the energy calculations can beimproved to the level of spin-orbit energies.

Ab initio calculations of crystal fields and crystal-field parameters have not yet reached the level of accu-racy that is required for most purposes. Progress in thatdirection is needed. Opportunities for theoretical re-search in the area abound.

From an experimental perspective, the issues ofunderstanding how interface anisotropy and magneto-striction can be manipulated and/or controlled will de-termine whether future devices can be developed. Forexample, what types of materials can be used at theinterface of a compositionally modulated layer to causeor eliminate perpendicular or in-plane anisotropy? Whatcan be done to make significantly softer magnetic films(NiFe, Co)? Can surface anisotropy be used to compen-sate bulk anisotropy in thin films? Can the dispersionin anisotropy be reduced by growing more perfectfilms? These very practical questions require a morethorough understanding, both theoretical and experi-mental, of many materials.

I. Magnetoresistivity

Galvanomagnetic effects in very pure materialshave been extensively used27 to determine very subtleproperties of their electronic structure, in particular thetopology of the Fermi surface in metals. The effects, onthe other hand, are more difficult to understand and in-terpret when macroscopic spatial heterogeneities play a


L. M. Falicov, D.T. Pierce ef a/.: Surface, interface, and thin-film magnetism

fundamental role. Such is the case with extended defects,shape and size imperfections, and surface roughness.

As mentioned previously (see Sec. II, TheoreticalBackground, Transport Properties in Magnetic Sys-tems), positive MR—an increase in the resistance uponapplication of a magnetic field—is associated with theconvoluted character of crystal electron orbits in a mag-netic field. Negative MR—a decrease in resistancewhen the field is applied—implies increased order, a re-duction in the strength of the electron scattering ora suppression of its sources. Negative MR in ferro-magnetic materials is qualitatively understood, and iscaused either by the removal of domain walls upon ap-plication of the field or by the introduction of a gap inthe spectrum of the spin waves, which are then less ef-fective in scattering the conduction electrons.

The recently discovered "giant" negative magneto-resistance in (001)Fe/(001)Cr superlattices96 can be aslarge as a factor of two, and has been attributed tothe spin-dependent transmission of conduction elec-trons between Fe layers through the Cr interlayers.This explanation is obviously only qualitative. Thephenomenon deserves considerable attention: (1) itsfundamental origins remain unclear, (2) it may perhapsbe found in systems more general than the specific onein which it was discovered, (3) it is most probably re-lated to the antiparallel coupling of the nearest neigh-bor Fe (individually ferromagnetic) layers, and (4) itwill almost certainly have important technological ap-plications (see Sec. V, Applications, A. 2: ExchangeBiasing of Magnetoresistive Heads).

J. Micromagnetics

The configuration of domains and domain walls isstrongly affected by the presence of surfaces or inter-faces at which there are demagnetizing fields and largecontributions to the magnetostatic energy. RecentSEMPA investigations34 have provided detailed insightinto how a Bloch wall in the bulk of a material like Feor permalloy turns over into a Neel wall at the sur-face. The Bloch wall can turn over into the Neel intwo different directions such that the walls are off-set. Micromagnetic calculations employing a continuumapproximation show quantitative agreement with ex-perimental asymmetric surface Neel wall profiles andoffsets. When two offset walls meet there is a topologi-cal singularity in the magnetization. The size of thecore of such a singularity remains a challenge to mea-sure with yet higher resolution SEMPA. A number ofsingularities have been discussed theoretically208'209 andinvite investigation with SEMPA. Other questions as tohow the magnetic microstructure changes as the speci-men thickness is decreased or what is the magneticstructure of one ferromagnetic film on top of anotherpresent further research opportunities.

A very interesting question is whether a film of asingle monolayer (if such could be realized in practice)would support domains. Recent work210 suggests that al-though uniform magnetization is expected for in-planemagnetization, domains are energetically favorable forperpendicular magnetization when the perpendicularsurface anisotropy exceeds a critical value. It would beof great interest to know how the size of the domainsand walls could be expected to vary in films of two,three, or more layers. It may be possible with SEMPAto investigate such domains and walls if suitable filmswith appropriate perpendicular anisotropy can be pre-pared. Further, an extension of the theory to finitetemperatures would allow direct comparison with ex-periment. In general, micromagnetic theory should beextended to the very small dimensions of these systemsso that calculation of domains and domain walls wouldbe possible. This would help clarify the relationship be-tween surface hysteresis loops measured with electronspectroscopies, and the bulk hysteresis curves charac-teristic of a deeper region.

At the microscopic level, domain walls may bezigzag instead of straight and sharp. If the domain wallis the transition between two written bits on a record-ing medium, the zigzag wall gives rise to noise and rep-resents an ultimate limitation on the recording density.Noise is also affected by the correlation of magnetiza-tion reversals for individual grains; this property maybe examined by experiments such as the anomalousHall effect.211 (These sources of noise may be reducedby decreasing the exchange coupling between grains.)With the push to ever higher density of informationstorage, there is much research needed to understand andcontrol magnetic microstructure at the microscopic level.

At the boundaries of a ferromagnet small closuredomains form to reduce the energy further. It is oftendesirable to control the closure domains. This can bedone by varying the shape of a magnetic element. Fur-ther, with a thin film, closure domains can be reducedby providing a return path for the edge flux through asecond layer. For example, a thin-film recording headmight consist of two magnetic layers separated by anonmagnetic layer of sufficient thickness to avoid ex-change coupling of the layers. Further development ofmeans to control the magnetic microstructure is facili-tated by new means of obtaining high-resolution imagesof the magnetization.

Domain nucleation is an issue which, if it wereunderstood, could provide considerable insight intoboth explaining experimental observations and design-ing new materials. Unfortunately, domain nucleation inreal materials probably involves a complicated relation-ship among the electronic structure near a defect, ther-mal fluctuations of the neighboring spins, and thenormal macroscopic forces of micromagnetics. An un-


L.M. Falicov, D.T. Pierce et at;. Surface, interface, and thin-film magnetism

usual opportunity to study this effect may be found incertain barium-ferrite particles which simultaneouslydisplay a high degree of crystalline perfection, a smallercoercivity than might be expected, a larger time depen-dence of coercivity than expected, and a thermally acti-vated magnetic dead layer on the surface.212'213 Areasonable conjecture is that the latter three propertiesare related and that fluctuations in the surface deadlayer are leading to domain nucleation and switching.Experimentally, a detailed characterization of the sur-face crystal structure might immediately lead to candi-date nucleation sites. It might, alternatively, provide afoundation for a detailed theoretical treatment whichcould include the local fluctuations balanced againstthe micromagnetic forces. It is worthwhile noting thatan alternative, although somewhat less easily character-ized system, is the y-Fe2O3 particle, where it is possiblethat Co surface doping removes nucleation sites.214

A predictive understanding of hysteresis loops, ofwhich the coercivity and remanent coercivity form par-ticularly interesting features, would be obviously bene-ficial. In some materials the coercivity and nucleationfield will coincide, but in many other cases nucleationwill not be the fundamental barrier. A fruitful ap-proach for some of these latter materials should be anextension of a currently available domain-wall pinningtheory31 toward greater quantitative accuracy. Calcula-tions of moment, exchange, and interface anisotropiesat grain boundaries will be particularly useful. Carefulexperimental study of select systems should provide thenature of the grain boundary, and studies of time de-pendence may help determine the shape of the energybarrier. In other materials, such as those exhibitingsharp, well-defined grains, further implementation ofthe approach of Ref. 32 would be appropriate. Herethe goal would be to compare the results of accurateimplementation of micromagnetic theory including,if necessary, domain nucleation, to equally accurateexperiments for a variety of systems beyond the CoNithin films originally treated. This will help determinethe physical limits of the theory.

In general, future advances will substantially de-pend on careful comparison among atomic structure,micromagnetic features observed by high spatial reso-lution techniques, and accurate computer simulations.Investigations will consider progressively smaller spa-tial scales.

K. Magnetics technology

Because of the many applications of magnetic ma-terials and phenomena, new insights into the basicphysics of magnetism often have technological implica-tions. An investigation of the fundamental properties ofthe interface, whether it be the surface-vacuum inter-face, the interface between thin film and substrate, or

the interface between magnetic layers, also provides anopportunity to contribute to the solution of many tech-nological problems. A deep understanding of the inter-actions at interfaces will allow scientists to controlmaterial properties. For example, a suitable underlayerfor a recording medium can control grain size and ori-entation. A fundamental understanding of interface de-coupling layers may, along with the ability to controlsegregation to grain interfaces, lead to reduced noisein magnetic recording media or increased coercivityin permanent-magnet materials. The current under-standing of interface-induced anisotropy or of exchangebiasing at an interface is insufficient for engineeringpurposes. Clearly, there are many research opportuni-ties into fundamental questions which also represent re-search opportunities in magnetics technology.

L. Conclusion

An effort has been made to describe briefly thecurrent status of research in surface, interface, and thinfilm magnetism and highlight some of the issues andresearch opportunities. Even though the discussion wasnecessarily brief, the report is lengthy owing to the di-versity of the field and the high and increasing level ofresearch activity. The assembled panel collectively hasa wide background in the subject area; nevertheless, itis impossible to be absolutely comprehensive in cover-age. Still other issues which offer interesting researchopportunities could have been discussed.

The research opportunities in surface, interface,and thin film magnetism are exciting and many. Ad-vances in the growth and preparation of magnetic sys-tems, an active research area itself, have led tomaterials with different crystal phases, altered latticeconstants, layered structures, and so on, in short, tonew magnetic systems. Characterization of these newmaterials has caused refinement of existing techniquesand development of new ones to determine structural,electronic, and magnetic properties. The theory of mag-netism in these lower dimensional systems, often aidedby the availability of powerful computational facilities,has been important in stimulating and understandingexperimental work. These factors jointly create espe-cially significant opportunities for research in this area.

Magnetism in bulk solids is a well-developed re-search area which has provided a fertile testing groundfor quantum mechanics, theories of many-body inter-actions and collective phenomena, and critical phenom-ena. Surfaces, interfaces, and thin films represent newmagnetic systems and are, further, building blocks formore complex systems such as multilayers. With thehelp of new materials technologies, these systems canbe prepared in metastable phases which have no bulkcounterpart. Size effects and lower dimensionality addinteresting new facets to the study of magnetic proper-



ties of surfaces, interfaces, and thin films. In short, wehave new materials, exhibiting new properties, and pre-senting many fascinating fundamental questions to beanswered.

Many of the developments in the study of mag-netism have been driven by requirements of magneticstechnology. This multibillion-dollar-a-year industryspans technologies from magnetic media for informa-tion storage to permanent magnets for motors. Creatinga new materials system and understanding its magneticproperties has the potential to make a significant con-tribution to technology. What may be fundamental re-search questions about interactions at interfaces mayultimately provide the information to control knowl-edgeably the coercivity and anisotropy in a thin film orthe exchange coupling between grains with a conse-

Table of acronyms.

ALS Advanced Light Source, Lawrence Berkeley LaboratoryAPS Advanced Photon Source, Argonne National LaboratoryCVD Chemical Vapor DepositionDPC Differential Phase ContrastECS Electron Capture SpectroscopyEXAFS Extended X-ray Absorption Fine StructureFEL Free-Electron LaserHEED High-Energy Electron DiffractionLEED Low-Energy Electron DiffractionMBE Molecular Beam EpitaxyMOCVD Metal-Organic Chemical Vapor DepositionMOKE Magneto-Optic Kerr EffectMR Magneto-Resistive or MagnetoResistanceMFM Magnetic Force MicroscopyNSLS National Synchrotron Light Source, Brookhaven

National LaboratoryPNR Polarized Neutron ReflectometryRE Rare EarthRHEED Reflection High-Energy Electron DiffractionRKKY Ruderman-Kittel-Kasuya-Yosida interactionSEM Scanning Electron MicroscopySEMPA Scanning Electron Microscope with Polarization

AnalysisSHG Second-Harmonic GenerationSMOKE Surface Magneto-Optic Kerr EffectSPEELS Spin-Polarized Electron Energy Loss SpectroscopySPLEED Spin-Polarized Low-Energy Electron DiffractionSPSEE Spin-Polarized Secondary-Electron EmissionSQUID Superconducting Quantum Interference DeviceSR Synchrotron RadiationSTEM Scanning Transmission Electron MicroscopySTM Scanning Tunneling MicroscopyTEM Transmission Electron MicroscopyTM Transition MetalUPS Ultraviolet Photoemission SpectroscopyVSM Vibrating Sample MagnetometerVUV Vacuum UltraVioletXPS X-Ray Photoemission Spectroscopy1-D One-Dimensional2-D Two-Dimensional3-D Three-Dimensional

quent impact on information storage devices. One ofthe exceptional aspects of magnetics research is thatprogress in fundamental issues and the solving of tech-nological problems often go hand in hand.

The research opportunities in surface, interface,and thin film magnetism touched on in this reportrange from basic to applied issues and include bothexperimental and theoretical questions which shouldchallenge researchers in university, government, andindustrial laboratories for a number of years. It is anexciting area to work in: there is much to be done.

REFERENCES

'V. L. Moruzzi, J. F. Janak, and A. R. Williams, Calculated Elec-tronic Properties of Metals (Pergamon, New York, 1978).

2H. J. F. Jansen, K. B. Hathaway, and A. J. Freeman, Phys. Rev. B30, 6117 (1984).

3E. P. Wolfarth, Ferromagnetic Materials, edited by E. P. Wolfarth(North Holland, Amsterdam, 1980), Vol. 1, p. 1.

4M. R. Norman and A. J. Freeman, in The Challenge of d and fElectrons, edited by D. R. Salahub and M. C. Zerner (AmericanChemical Society, Washington, DC, 1989), p. 273.

5C. L. Fu, A. J. Freeman, and T. Oguchi, Phys. Rev. Lett. 54, 2700(1985); R. Richter, J.G. Gay, and J. R. Smith, Phys. Rev. Lett.54, 2704 (1985).

6See, for example, D. Wang, A. J. Freeman, and M. Weinert,J. Magn. Magn. Mat. 31-34, 891 (1983).

7S. C Hong, A. J. Freeman, and C. L. Fu, J. de Physique C8, 1683(1988).

8S. S. Peng and H. J. F. Jansen, J. Appl. Phys. 64, 5607 (1988).9C. S. Wang, B. M. Klein, and H. Krakauer, Phys. Rev. Lett. 54,1852 (1985).

10K. B. Hathaway, H. J. F. Jansen, and A. J. Freeman, Phys. Rev. B31, 7603 (1985).

"C. S. Wang and J. Callaway, Phys. Rev. B 15, 298 (1977).12T. Jarlborg and M. Peter, J. Magn. Magn. Mat. 42, 89 (1984).13A discussion of some discrepancies between theory and de

Haas-van Alphen data for Co may be found in F. J. Himpsel andD. E. Eastman, Phys. Rev. B 21, 3207 (1980).

14See, for example, R. H. Victora, in Magnetic Properties of Low-Dimensional Systems, edited by L.M. Falicov and J. L. Moran-Lopez (Springer-Verlag, Berlin-Heidelberg-New York-Tokyo,1986), p. 25.

15R. Car and M. Parinello, Phys. Rev. Lett. 55, 2471 (1985); ibid.60, 204 (1988).

16H. Brooks, Phys. Rev. 58, 909 (1940).17M.T. Hutchings, in Solid State Physics, edited by F. Seitz and

D. Turnbull (Academic Press, New York, 1964), Vol. 16, p. 227,and references therein.

18L. Neel, J. Phys. Rad. 15, 225 (1954).19D.C. Mattis, The Theory of Magnetism (Harper and Row,

New York, 1965), Chap. 9.20C. N. Yang, Phys. Rev. 85, 809 (1952).21M. Toda, R. Kubo, and N. Saito, Statistical Physics I (Equilib-

rium Statistical Mechanics) (Springer-Verlag, Berlin-Heidelberg-New York-Tokyo, 1983), p. 149.

22H. E. Stanley, Introduction to Phase Transitions and Critical Phe-nomena (Oxford University Press, Oxford, 1971).

23J. M. Kosterlitz and D. J. Thouless, in Progress in Low Tempera-ture Physics, edited by D. F. Brewer (North Holland, Amsterdam,1978), Vol. VII B.



24M. E. Fisher and A. E. Ferdinand, Phys. Rev. Lett. 19, 169(1967).

^R. Lipowsky, Phys. Rev. Lett. 49, 1575 (1982); R. Lipowsky andW. Speth, Phys. Rev. B. 28, 2983 (1983); R. Lipowsky, Z. Phys. B55, 345 (1984); R. Lipowsky, Phys. Rev. Lett. 52, 1429 (1984);R. Lipowsky, Phys. Rev. B 32, 1731 (1985); R. Lipowsky, Ferro-electrics 73, 69 (1987); R. Lipowsky, in Random Fluctuationsand Pattern Growth, edited by H. E. Stanley and N. Ostrowsky(Kluwer Academic, Dordrecht, 1988).

26S. Dietrich, in Phase Transitions and Critical Phenomena, editedby C. Domb and J. L. Leibowitz (Academic Press, London,1988), Vol. 8, p. 1.

27See, for instance, J. M. Ziman, Principle of the Theory of Solids(Cambridge, 1972), 2nd ed., pp. 250ff, 301ff; A. B. Pippard, Mag-netoresistance in Metals (Cambridge University Press, Cambridge,1989).

^G. R. Taylor, A. Isin, and R.V. Coleman, Phys. Rev. 165, 621(1968); R.V. Coleman, R.C. Morris, and D.J. Sellmyer, Phys.Rev. B 8, 317 (1973); R.W. Klaffky and R.V. Coleman, Phys.Rev. B 10, 4803 (1974).

29G. G. Cabrera and L. M. Falicov, Phys. Status Solidi B 61, 539(1974); ibid., 62, 217 (1974); G. G. Cabrera and L. M. Falicov,Phys. Rev. B 11, 2651 (1975).

30W. F. Brown, Micromagnetics (Wiley, New York, 1963).31R. Friedberg and D.I. Paul, Phys. Rev. Lett. 34, 1234 (1975).32R. H. Victora, Phys. Rev. Lett. 58, 1788 (1987).33N. Smith, J. Appl. Phys. 63, 2932 (1988); M.R. Scheinfein, J.

Unguris, R. J. Celotta, and D.T. Pierce, Phys. Rev. Lett. 63, 668(1989).

34See, for instance, K.E. Meyer, I.K. Schuller, and C M . Falco,J. Appl. Phys. 52, 5803 (1981).

35T. Venkatesan, X.D. Wu, B.D. Dutta, A. Inam, M.S. Hedge,D. M. Hwang, C. C. Chang, L. Nazar, and B. Wilkens, in Scienceand Technology of Thin Film Superconductors, edited by R. D.McConnell and S. A. Wolf (Plenum, New York, 1988), p. 1.

36R. Kaplan, J. Vac. Sci. Tech. Al, 551 (1983).37R.E. Ceck and E.I. Alessandrini, Trans. Am. Soc. Met. 51, 50

(1959).38See, for instance, Epitaxial Growth, edited by J.W. Matthews

(Academic Press, New York, 1975).39R.W. Erwin, J. J. Rhyne, M. B. Salamon, J. Borchers, S. Sinha,

R. Du, J. E. Cunningham, and C. P. Flynn, Phys. Rev. B 35, 6808(1987).

""For an extensive compilation of epitaxial systems see, for in-stance, E. Grunbaum, Ref. 38, p. 611.

41See, for instance, K. P. Staudhammer and O. E. Meur, Atlas ofBinary Alloys (Marcel Dekker, New York, 1973).

42See various articles in Physics, Fabrication and Applications ofMultilayered Structures, edited by P. Dhez and C. Weisbuch(Plenum Press, New York, 1988).

43S. A. Chambers, J. J. Wagener, and J. H. Weaver, Phys. Rev. B36, 8992 (1987).

*TXA. Steigerwald and W.F. Egelhoff, Jr., Surf. Sci. 192, L887(1987).

45W. A. A. Macedo and W. Keune, Phys. Rev. Lett. 61, 475 (1988)."C. Liu, E. R. Moog, and S. D. Bader, Phys. Rev. Lett. 60, 2422

(1988).47M. Stampanoni, A. Vaterlaus, M. Aeschlimann, F. Meier, and D.

Pescia, J. Appl. Phys. 64, 5321 (1988).^C. Liu and S.D. Bader, Physica B 161, 253 (1989).49S. Blugel, M. Weinert, and P. H. Dedericks, Phys. Rev. Lett. 60,

1077 (1988).50G. Allan, Surf. Sci. 74, 79 (1978).51D. R. Grempel, Phys. Rev. B 24, 3928 (1981).52R. H. Victora and L. M. Falicov, Phys. Rev. B 31, 7335 (1985).53G. A. Prinz, Phys. Rev. Lett. 54, 1051 (1985).

54T. Takahashi and W. A. Bassett, Science 145, 483 (1964).55V.L. Moruzzi, P.M. Marcus, K. Schwarz, and P. Mohn, Phys.

Rev. B 34,1784 (1986); P. M. Marcus, V. L. Moruzzi, Z. Q. Wang,Y. S. Li, and F. Jena, in Physical and Chemical Properties of ThinMetal Overlayers and Alloy Surfaces (Proc. Mater. Res. Soc.Symp.), edited by D. M. Zener and D.W. Goodman (MaterialsResearch Society, Pittsburgh, PA, 1987), Vol. 83, p. 21.

56G.A. Prinz, B.T. Jonker, J. J. Krebs, J. M. Ferrari, and F.Kovanis, Appl. Phys. Lett. 48, 1756 (1986).

57M.W. Ruckman, J. J. Joyce, and J. J. Weaver, Phys. Rev. B 33,7029 (1986).

58D. Weller and S. F. Alvarado, J. Appl. Phys. 59, 2908 (1986).59J. Kwo, D. B. McWhan, M. Hong, E. M. Gyorgy, L. C. Feldman,

and J. E. Cunningham, in Layered Structures, Epitaxy and Inter-faces (Proc. Mater. Res. Soc. Symp.), edited by J. M. Gibson andL. R. Dawson (Materials Research Society, Pittsburgh, PA,1985), Vol. 37, p. 509.

'"H. Homma, K. Yang, and I. K. Schuller, Phys. Rev. B 36, 9435(1987).

61J. J. Rhyne, R.W. Erwin, M. B. Salamon, S. Sinha, J. Borchers,J. E. Cunningham, and C. P. Flynn, J. Less-Common Metals 148,17 (1989).

62C.F. Majkrzak, D. Gibbs, P. Boni, A.I. Goldman, J. Kwo, M.Hong, T. C. Hsieh, R. M. Fleming, D. B. McWhan, J.W. Cable,J.Bohr, H. Grimm, and C.L. Chien, J. Appl. Phys. 63, 3447(1988).

63J. Kwo, in Thin Film Growth Techniques for Low-DimensionalStructures, edited by R.F.C. Farrow, S.S.P. Parkin, P. J. Dob-son, J. H. Neave, and A. S. Arrott, NATO ASI Series B (Plenum,New York, 1987), Vol. 163, p. 337.

"R. DU, F. Tsui, and C. P. Flynn, Phys. Rev. B 38, 2941 (1988).65C.P. Flynn, F. Tsui, M. B. Salamon, R.W. Erwin, and J. J.

Rhyne, J. Phys.: Condensed Matter 1, 5997 (1989).66See, for instance, L. A. Bruce and H. Jaeger, Phil. Mag. A 38,

223 (1978).67K.Y. Yang, H. Homma, and I. K. Schuller, J. Appl. Phys. 63, 4066

(1988).68See various articles in Mater. Res. Soc. Bull. XIV (1989).69A. E. Berkowitz and J. H. Grenier, J. Appl. Phys. 36, 3330 (1965);

J. Bransky, I. Bransky, and A. A. Hirsch, J. Appl. Phys. 41, 183(1970).

70K. B. Alexander, F. J. Walker, R. A. McKee, and F. A. List, sub-mitted to the Am. Ceram. Soc; R.A. McKee, F. A. List, F. J.Walker, and J.R. Conner, submitted to Phys. Rev. B.

7IA. Berkowitz (private communication).72G. A. Prinz, Mater. Res. Soc. Bull. XIII, 2 (1988)."S. D. Bader, in Proc. IEEE, Special Issue on Magnetics, edited by

R. M. White (IEEE, New York, 1990), in press.74R. H. Victora and L. M. Falicov, Phys. Rev. B 30, 259 (1984).75A. J. Freeman and C. L. Fu, in Magnetic Properties of Low-Dimensional Systems, edited by L. M. Falicov and J. L. Moran-Lopez (Springer-Verlag, Berlin-Heidelberg-New York-Tokyo,1986), p. 16.

76D. S. Wang, A. J. Freeman, and H. Krakauer, Phys. Rev. B 24,1126 (1981).

77L.M. Falicov, R. H. Victora, and J. Tersoff, in The Structure ofSurfaces, edited by M. A. Van Hove and S.Y. Tong (Springer-Verlag, Berlin-Heidelberg-New York-Tokyo, 1985), p. 12, andreferences therein.

78R.H. Victora and L.M. Falicov, Phys. Rev. B 31, 7335 (1985).79L. E. Klebanoff, R. H. Victora, L. M. Falicov, and D. A. Shirley,

Phys. Rev. B 32, 1997 (1985).*°J. Tersoff and L. M. Falicov, Phys. Rev. B 26, 6186 (1982).81R. H. Victora, S. Ishida, and L. M. Falicov, Phys. Rev. B 30,3896

(1984).82H. Hasegawa, J. Phys. F: Met. Phys. 16, 1555 (1986).



83W. H. Meiklejohn and C. P. Bean, Phys. Rev. B 102, 1413 (1959).84See, for example, W. Stoecklin, S.S. P. Parkin, and J. C. Scott,

Phys. Rev. B 38, 6854 (1988), and references therein.85A. Yelon, in Physics of Thin Films, edited by M. Francombe and

R. Hoffmann (Academic Press, New York, 1971), Vol. 6, p. 205.86A. P. Malozemoff, Phys. Rev. B 35, 3679 (1987); J. Appl. Phys.

63, 3874 (1988).87B. Hermsmeier, J. Osterwalder, D. Friedman, and C. S. Fadley,

Phys. Rev. Lett. 62, 478 (1989).88S. S. P. Parkin and V. S. Speriosu, in Magnetic Properties of Low-

Dimensional Systems, Proc. of the 2nd Int. Workshop held in SanLuis Potosf, Mexico, May 22-26, 1989, edited by L. M. Falicovand J. L. Mofan-Lopez (Springer-Verlag, Heidelberg-Berlin-NewYork-Tokyo, 1989) (in press).

89B. Heinrich, S.T. Purcell, J. R. Dutcher, K. B. Urquhart, J. F.Cochran, and A. S. Arrott, Phys. Rev. 38, 12879 (1988).

90R. Nakatani, T. Kobayashi, S. Ootomo, and N. Kumusaka, Jpn.J. Appl. Phys. 27, 937 (1988).

91M. Aeschlimann, Ph.D. Dissertation, ETH 8863 (1989); M.Aeschlimann, G. L. Bona, F. Meier, M. Stampanoni, A. Vater-laus, H. C. Siegmann, E. E. Marinero, and H. Notarys, IEEETrans. Mag. MAG-24, 3180 (1988).

92W. C. Cain, J.W. Lee, P.V. Koeppe, and M. H. Kryder, IEEETrans. Mag. MAG-24, 2609 (1988).

93T. Kobayashi, H. Tsuji, S. Tsunashima, and S. Uchiyama, Jpn. I.Appl. Phys. 20, 2089 (1981).

94C. Carbone and S. F. Alvarado, Phys. Rev. B 36, 2433 (1987).95P. Griinberg, R. Schreiber, Y. Pang, M. B. Brodsky, and H.

Sowers, Phys. Rev. Lett. 57, 2442 (1986); F. Saurenbach, U. Walz,L. Hinchey, P. Griinberg, and W. Zinn, J. Appl. Phys. 63,3473 (1988).

96M. N. Baibich, J. M. Broto, A. Fert, F. Nguyen Van Dau, F.Petroff, P. Etienne, G. Creuzet, A. Friederich, and J. Chazelas,Phys. Rev. Lett. 61, 2472 (1988).

97B.T. Jonker, K. H. Walker, E.Kisker, G.A. Prinz, and C. Car-bone, Phys. Rev. Lett. 57, 142 (1986).

98C.J. Gutierrez, S. H. Mayer, Z.Q. Qiu, H. Tang, and J. C.Walker, in Growth, Characterization and Properties of UltrathinMagnetic Films and Multilayers (Proc. Mater. Res. Soc. Symp.),edited by B.T. Jonker, E. E. Marinero, and J. P. Heremans (Ma-terials Research Society, Pittsburgh, PA, 1989), Vol. 151, p. 17.

"J. Mathon, Rep. Prog. Phys. 51, 1 (1988).100J. Mathon and S. B. Ahmad, Phys. Rev. B 37, 660 (1988);

J. Mathon, Physica 149B, 31 (1988).101D. Mauri, D. Scholl, H. C. Siegmann, and E. Kay, Phys. Rev.

Lett. 62, 1900 (1988).102J. S. Moodera and R. Meservey, Phys. Rev. B 34, 379 (1986).103J.C. Slonczewski, Phys. Rev. B 39, 6995 (1988).104M. Julliere, Phys. Lett. 54A, 225 (1975).10SP.M. Tedrow and R. Mesetvey, Phys. Rev. B 7, 318 (1973).106U. Gaefvert and S. Maekawa, IEEE Trans. Mag. MAG-18, 707

(1982).107C. F. Majkrzak, J.W. Cable, J. Kwo, M. Hong, D. B. McWhan, Y.

Yafet, J.V. Waszczak, and C. Vettier, Phys. Rev. Lett. 56, 2700(1986).

™M.B. Salamon, S. Sinha, J. J. Rhyne, J.E. Cunningham, R.W.Erwin, J. Borchers, and C.P. Flynn, Phys. Rev. Lett. 56, 259(1986).

109Y. Yafet, J. Kwo, M. Hong, C. F. Majkrzak, and T. O'Brien, J.Appl. Phys. 63, 3453 (1988).

"°J. A. Borchers, G. Nieuwenhuys, M.B. Salamon, C.P. Flynn, R.Du, R.W. Erwin, and J. J. Rhyne, J. de Physique 49-C3, 1685(1988).

mSee, for instance, I. K. Schuller and H. Homma, Mater. Res. Soc.Bull. XII, 1 (1987).

m C . Colvard, R. Merlin, M.V. Klein, and A. C. Gossard, Phys.

Rev. Lett. 45, 298 (1980).'"See, for instance, J. Sokoloff, Solid State Commun. 40, 633

(1981)."4A. P. van Gelder, Phys. Rev. 181, 7887 (1969).U5R.E. Camley, T.S. Rahman, and D. Mills, Phys. Rev. B 27, 261

(1983).mA. Kueny, M. R. Khan, I. K. Schuller, and M. Grimsditch, Phys.

Rev. B 29, 2879 (1984)."'"Panel Report on Fundamental Issues in Heteroepitaxy" Mon-

terey, CA, January 1989 (to be published).1I8P. J. Flanders, J. Appl. Phys. 63, 3940 (1988).U9E. R. Moog and S. D. Bader, Superlatt. Microstr. 1, 543 (1985).™K. B. Urquhart, B. Heinrich, J. F. Cochran, A. S. Arrott, and K.

Myrtle, J. Appl. Phys. 64, 5335 (1988).121M. Farle and K. Baberschke, Phys. Rev. Lett. 58, 511 (1987).122Th. Frey, W. Jantz, and S. Stibal, J. Appl. Phys. 64, 6002 (1988).123See D. S. Bloomberg and G. A. N. Connell, in Magnetic Record-

ing, Vol. II: Computer Data Storage, edited by C. D. Mee and E. D.Daniel (McGraw-Hill, New York, 1988).

124F. Schmidt, W. Rave, and A. Hubert, IEEE Trans. Magn. MAG-21, 1596 (1985).

125D. A. Herman, Jr. and B.E. Argyle, IEEE Trans. Magn. MAG-22, 772 (1986).

126C. Liu and S. D. Bader, in Magnetic Properties of Low Dimen-sional Systems II, edited by L. M. Falicov, F. Mejia-Lira, and J. L.Moran-Lopez (Springer-Verlag, Berlin, 1990), p. 22.

127W. Reim, J. Magn. Magn. Mat. 58, 1 (1986).128W. Hiibner and K. H. Bennemann, "Nonlinear Magneto-Optical

Kerr Effect at the Nickel Surface," preprint submitted forpublication.

129Y. R. Shen, in Chemistry and Structure at Interfaces: New Laserand Optical Techniques, edited by R. B. Hall and A. B. Ellis (Ver-lag Chemie, Weinheim, 1986), p. 151.

130B. Hillebrands, P. Baumgart, and G. Giintherodt, Phys. Rev. B36, 2450 (1987).

13IM. Grimsditch, A. Malozemoff, and A. Brunsch, Phys. Rev.Lett. 43, 711 (1979).

132P. Griinberg, R. Schreiber, Y. Pang, M.B. Brodsky, and H.Sowers, Phys. Rev. Lett. 57, 2442 (1986).

133B. Heinrich, A. S. Arrott, J. F. Cochran, C. Liu, and K. Myrtle,J. Vac. Sci. Technol. A 4, 1376 (1986).

134M. Campagna, D.T. Pierce, F. Meier, K. Sattler, and H. C.Siegmann, Adv. Electronics and Electron Phys. 41, 113 (1976).

135E. Kisker, K. Schroeder, W. Gudat, and M. Campagna, Phys.Rev. B 31, 329 (1985).

136P. D. Johnson, A. Clarke, N. B. Brookes, S.L. Hulbert, B.Sinkovic, and N.V. Smith, Phys. Rev. Lett. 61, 2257 (1988).

137D. Weller, S. F. Alvarado, W. Gudat, K. Schroder, and M.Campagna, Phys. Rev. Lett. 54, 1555 (1985).

138C. Carbone and E. Kisker, Solid State Commun. 65, 1107 (1988).139Y. Kakehashi, Phys. Rev. B 31, 7482 (1985).140M. Taborelii, R. Allenspach, G. Boffa, and M. Landolt, Phys.

Rev. Lett. 56, 2869 (1986).141M.A. Van Hove and S.Y. Tong, Surface Crystallography

(Springer-Verlag, Berlin-Heidelberg-New York-Tokyo, 1979).142For an overview of spin-polarized electron spectroscopy tech-

niques see, for example, Polarized Electrons in Surface Physics,edited by R. Feder (World Scientific, Singapore, 1985).

143D.T. Pierce, R. J. Celotta, J. Unguris, and H. C. Siegmann, Phys.Rev. B 26, 2566 (1982).

144S. F. Alvarado, M. Campagna, and H. Hopster, Phys. Rev. Lett.48, 51 (1982).

145W. Diirr, M. Taborelii, O. Paul, R. Germar, W. Gudat, D. Pescia,and M. Landolt, Phys. Rev. Lett. 62, 206 (1989).

146See the article by U. Gradmann and S. F. Alvarado in Ref. 141.147D.T. Pierce, R. J. Celotta, G-C. Wang, W. N. Unertl, A. Galejs,



C.E. Kuyatt, and S.R. Mielczarek, Rev. Sci. Instrum. 51, 478(1980).

148D. R. Penn, S. P. Apell, and S. M. Girvin, Phys. Rev. B 32, 7753(1985).

149D. L. Abraham and H. Hopster, Phys. Rev. Lett. 58, 1352 (1987).150D. Mauri, D. Scholl, H. C. Siegmann, and E. Kay, Phys. Rev.

Lett. 62, 1900 (1989).151A. Venus and J. Kirscher, Phys. Rev. B 37, 2199 (1988); Y.U.

Idzerda, D. M. Lind, D. A. Papaconstantopoulos, G.A. Prinz,B.T. Jonker, and J.J. Krebs, Phys. Rev. Lett. 61, 1222 (1988);D. L. Abraham and H. H. Hopster, Phys. Rev. B 62, 1157 (1989).

152D. L. Abraham, Ph.D. Thesis, University of California, Irvine,CA (1989) (unpublished research).

153M. Onellion, M.W. Hart, F. B. Dunning, and G.K. Walters,Phys. Rev. Lett. 52, 380 (1984).

154C. Rau, C. Schneider, G. Xing, and K. Jamison, Phys. Rev. Lett.57, 3221 (1986).

155D.W. Gidley, A.R. Koymen, and T.W. Capehart, Phys. Rev.Lett. 49, 1779 (1982).

15(>J. Unguris, D.T. Pierce, and R. J. Celotta, Rev. Sci. Instrum. 57,1314 (1986).

157G. G. Hembree, J. Unguris, R. J. Celotta, and D.T. Pierce, Scan-ning Microscopy Supplement 1, 229 (1987).

158D.T. Pierce, J. Unguris, and R. J. Celotta, Mater. Res. Soc. Bull.13, 19 (1988).

159M. R. Scheinfein, D.T. Pierce, J. Unguris, J. J. McClelland, R. J.Celotta, and M.H. Kelley, Rev. Sci. Instrum. 60, 1 (1989).

i60D T P i e r c e ( M R Scheinfein, J. Unguris, and R. J. Celotta, inGrowth, Characterization and Properties of Ultrathin MagneticFilms and Multilayers (Proc. Mater. Res. Soc. Symp.), edited byB.T. Jonker, E. E. Marinero, and J. P. Heremans (Materials Re-search Society, Pittsburgh, PA, 1989), Vol. 151, p. 49.

161D. C. Joy and J. P. Jacubovics, J. Phys. D 2, 1367 (1969).162A. Tonomura, T. Matsuda, J. Endo, T. Arii, and K. Mihama,

Phys. Rev. Lett. 44, 1430 (1980).163N. H. Dekkers and H. deLang, Optik 41, 452 (1974).i aJ. N. Chapman and G. R. Morrison, J. Magn. Magn. Mat. 35, 254

(1983).165K. Tsuno, Rev. Solid State Sci. 2, no. 4, 623 (1988).166D.T. Pierce, Physica Scripta 38, 291 (1988).167Y. Martin, D. Rugar, and H. K. Wiskramasinghe, Appl. Phys.

Lett. 52, 244 (1988).168U. Hartmann, J. Appl. Phys. 64, 1561 (1988).169For recent reviews see, for instance, the various chapters in Inter-

faces, Superlattices and Thin Films (Proc. Mater. Res. Soc. Symp.),edited by J. D. Dow and I. K. Schuller (Materials Research Soci-ety, Pittsburgh, PA, 1987), Vol. 77.

™See, for instance, D. B. McWhan, chapter in Ref. 169.171See, for instance, E. Spiller, chapter in Ref. 169.172See, for instance, various articles in Multilayers: Synthesis, Prop-

erties and Non-Electronic Application (Proc. Mater. Res. Soc.Symp.), edited by T.W. Barbee, Jr., F. Spaepen, and L. Greer(Materials Research Society, Pittsburgh, PA, 1988), Vol. 103.

173See, for instance, J. P. Locquet, D. Neerink, W. Sevenhans,Y. Bruynseraede, H. Homma, and I. K. Schuller, chapter inRef. 172, p. 211.

I74I. K. Schuller, Y. Bruynseraede, E. Fullerton, and H. Vander-straeten (to be published).

175J. J. Rhyne, R.W. Erwin, J. Borchers, M. B. Salamon, R. Du, andC. P. Flynn, Physica B 159, 111 (1989).

176G. P. Felcher, R. O. Hilleke, R. K. Crawford, J. Hanmann, R.Kleb, and G. Ostowski, Rev. Sci. Instrum. 58, 609 (1987).

I77S.S.P. Parkin, R. Sigsbee, R. Felici, and G.P. Felcher, Appl.Phys. Lett. 48, 604 (1986).

178M. F. Toney, T. C. Huang, S. Brennan, and Z. Rek, J. Mater.Res. 3, 351 (1988).

179S.S.P. Parkin, V. Deline, R. Hilleke, and G.P. Felcher, privatecommunication and submitted to Phys. Rev. B.

180J. A. C. Bland and R. F. Willis, in Thin Film Growth Techniquesfor Low-Dimensional Structures, edited by R. F. C. Farrow, S. S. P.Parkin, P. J. Dobson, J. H. Neave, and A. S. Arrott, NATO ASISeries B (Plenum, New York, 1987), Vol. 163, p. 405.

1S1H. Zabel (private communication).182W. Eberhardt and C.S. Fadley, "Summary of the Workshop on

New Opportunities in Surface Science at the Advanced LightSource," Lawrence Berkeley Laboratory Publication 5215, Sep-tember 1988.

183C. A. Brau, Science 239, 1115 (1988).wiFree Electron Laser Applications in the Ultraviolet (Optical Society

of America, Washington, DC, 1988).185See S.D. Bader, in Ref. 184, p. 20.186S. D. Bader, in Proc. of the First Users Meeting for the Advance

Photon Source, Argonne National Laboratory, APS-CP-1,November 1986.

187B. Sinkovic, B. Hermsmeier, and C. S. Fadley, Phys. Rev. Lett.55, 1227 (1985).

188W.C. Marra, P. Eisenberger, and A.Y. Cho, J. Appl. Phys. 50,6927 (1979); W. C. Marra, P. H. Fuoss, and P. Eisenberger, Phys.Rev. Lett. 49, 1169 (1982).

189D. Gibbs, D. E. Moncton, K.L. DAmico, J. Bohr, and B. H.Grier, Phys. Rev. Lett. 55, 234 (1985).

190C. Vettier, D. B. McWhan, E. M. Gyorgy, J.R. Kwo, B. M.Buntschuh, and B.W. Batterman, Phys. Rev. Lett. 56, 757 (1985).

191M. Blume, J. Appl. Phys. 57, 3615 (1985).192D.J. Rogers, J. N. Chapman, J. P.C. Bernards, and S. B.

Luitjens, IEEE Trans. Mag. MAG-25, 4180 (1989); K. Hono andD. E. Laughlin, J. Magn. Magn. Mater. 80, L137 (1989).

I93T. Lin, R. Alani, and D. N. Lambeth, J. Magn. Magn. Mater. 78,213 (1989).

194J. A. Christner, R. Ranjan, R. L. Peterson, and J. I. Lee, J. Appl.Phys. 63, 3260 (1988).

195B. R. Natarajan and E. S. Murdock, IEEE Trans. Mag. MAG-24,2724 (1988).

196W. B. Zeper, F. J. A. M. Greidanus, P. F. Carcia, and C. R.Fincher, J. Appl. Phys. 65, 4971 (1989).

197W. C. Cain, D. C. Markham, and M. H. Kryder, IEEE Trans.Mag. MAG-25, 3695 (1989).

198F. J. Cadieu, H. Hedge, and K. Chen, IEEE Trans. Mag. MAG-25, 3788 (1989).

199W. E. Wallace, Progress in Solid State Chem. 16, 127 (1985).200P. Bagno, O. Jepsen, and O. Gunnarson, Phys. Rev. Lett. 40,

1997 (1989).^ C . Liu and S. D. Bader, Phys. Rev. B 41, 553 (1990).202D. Mauri, D. Scholl, H. C. Siegmann, and E. Kay, Phys. Rev.

Lett. 61, 758 (1988).^G.T. Rado, Bull. Am. Phys. Soc. 2, 127 (1957).204D. L. Mills and A. A. Maradudin, J. Phys. Chem. Solids 28, 1855

(1967).205P. Grunberg and K. Mika, Phys. Rev. B 27, 2955 (1983); K. Mika

and P. Grunberg, Phys. Rev. B 31, 4465 (1985).206M. Grimsditch, M. R. Khan, A. Kueny, and I. K. Schuller, Phys.

Rev. Lett. 51, 498 (1983).^ A . Morrison, M. H. Kang, and E. J. Mele, Phys. Rev. B 39, 1575

(1989).208A. S. Arrott, B. Heinrich, and A. Aharoni, IEEE Trans. Magn.

MAG-15, 1228 (1979).^R . Vlaming and H. A.M. van den Berg, J. Appl. Phys. 63, 4330

(1988).210Y. Yafet and E. M. Gyorgy, Phys. Rev. B 38, 9145 (1988).211B. C. Webb and S. Schultz, IEEE Trans. Mag. MAG-24, 3006

(1988).212O. Kubo, T. Ido, H. Yokoyama, and Y. Koike, J. Appl. Phys. 57,


L.M. Falicov, D.T. Pierce ef a/.: Surface, interface, and thin-film magnetism

4280(1985). 214A.E. Berkowitz, F. E. Parker, E.L. Hall, and G. Podolsky,213R. H. Victora, J. Appl. Phys. 63, 3423 (1988); Phys. Rev. Lett. 63, IEEE Trans. Mag. MAG-24, 2871 (1988).

457 (1989).


Date post:	26-Mar-2020
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

MATERIALS REPORTS Surface, interface, and thin-film magnetismischuller.ucsd.edu/pdfs/Paper...

Documents