40TH ANNIVERSARY

J M A T E R S C I 4 1 (2 0 0 6 ) 7 9 3 –8 1 5

Nanomagnetism and spin electronics: materials,

microstructure and novel properties

K. M. KRISHNAN ∗, A. B. PAKHOMOV, Y. BAO, P. BLOMQVIST, Y. CHUN,M. GONZALES, K. GRIFFIN, X. JI , B. K. ROBERTSDepartment of Materials Science and Engineering, University of Washington, Box 352120,Seattle, WA, 98195-2120E-mail: kannanmk@u.washington.edu

We present an overview of the critical issues of current interest in magnetic andmagnetoelectronic materials. A broad demonstration of the successful blend of materialssynthesis, microstructural evolution and control, new physics and novel applications that iscentral to research in this field is presented. The critical role of size, especially at the nanometerlength scale, dimensionality, as it pertains to thin films and interfaces, and the overallmicrostructure, in determining a range of fundamental properties and potential applications, isemphasized. Even though the article is broad in scope, examples are drawn mainly from ourrecent work in magnetic nanoparticles and core-shell structures, exchange, proximity andinterface effects in thin film heterostructures, and dilute magnetic dielectrics, a new class ofmaterials for spin electronics applications. C© 2006 Springer Science + Business Media, Inc.

1. IntroductionMagnetism, subtle in its manifestations, is electronicallydriven but weak compared with electrostatic interactions.It has its origin in quantum mechanics, i.e. the Pauli ex-clusion principle and the existence of electron spin. How-ever, it is also known for a variety of classical effects,arising from both short- and long-range forces and iswidely associated with “microstructure” or the morpho-logical arrangements of phases, grains or individual atomsthemselves. These are, in part, the reasons for the rich-ness of structures and properties encountered in magneticsystems from which various useful and technological ap-plications arise [1]. These are also, in part, the reasonswhy the collective magnetic behavior of materials is com-plex, poorly understood and many fundamental questionsremain unanswered.

There are clearly two limits to magnetic behavior asa function of size and dimensionality. At one end of thespectrum (bulk) the microstructure determines the mag-netic (hard and soft) behavior. It is a function of the pro-cessing method and our understanding of it is qualitativeand empirical at best. At the other end, as the length scalesapproach the size of domain wall-widths (nanostructures),lateral confinement (shape and size) and inter-particleexchange effects dominate, rendering classical descrip-tions grossly inadequate, until finally, at atomic dimen-

∗Author to whom all correspondence should be addressed.

sions quantum-mechanical tunneling effects are expectedto predominate [2].

In the ideal case, considering only dipolar interactions,the spin-flip barrier for a small magnetic object [3] isa product of the square of the saturation magnetization,Ms and its volume (V ∼ a3). As a first approximation,one can set this equal to the thermal energy, i.e. Ms

2 a3

∼ kBT ∼ 25 meV at room temperature, and for typicalferromagnets obtain a characteristic length a ∼ 4 nm, be-low which they do not behave as ferromagnets any more.In practice, changes in magnetization of a material oc-cur via activation over an energy barrier and associatedwith each type of energy barrier is a different physicalmechanism and a characteristic length. These fundamen-tal lengths are the crystalline anisotropy length (lK∼(J/K),the magnetostatic length (ls∼

√J/(2π ·Ms

2)) and the ap-plied field length (lH∼(2J/H·Ms). Here J is the inter-atomic exchange, K is the anisotropy constant of the bulkmaterials and H the applied field. In principle, if multi-ple barriers are present, for a given time, the one withthe shortest characteristic length determines the materialsproperties [4]. For a general anisotropy, K, a characteristictime, τ , for reversal is determined as τ∼τ 0 exp (K·V/kBT).If the measurement time (typically 100 s) is included,one can then determine a characteristic size (Vsp) atroom temperature or, for a given volume, a characteristic

0022-2461 C© 2006 Springer Science + Business Media, Inc.DOI: 10.1007/s10853-006-6564-1 793

40TH ANNIVERSARY

Figure 1 Single domain size, Dcrit and magnetic stability size or the super-paramagnetic limit at room temperature, Dsp for some common ferromag-netic materials.

temperature called the blocking temperature, Tb that de-fines a transition from ferromagnetic to a thermally unsta-ble or superparamagnetic behavior. Further, if the particlesizes become larger, instead of being uniformly magne-tized (single domain), the particles can break into multipledomains to minimize their overall energy. Using theoriesfor domain stability in fine particles [5] and bulk prop-erties available in the literature, one can determine thecharacteristic size up to which single domains are stable.This series of magnetic “phases” as a function of size isshown (Fig. 1) for different ferromagnets and includes a“single domain” size (Dcrit) below which the material willnot support a multi-domain particle [6] and a size (Dsp)defined by the super-paramagnetic effect [7] below whicha spontaneous flip in magnetization occurs due to thermaleffects at room temperature.

Such nanostructures in the size range identified inFig. 1, which can be considered to be zero-dimensionalobjects, can be made by three distinct approaches – met-allurgy, chemical synthesis and lithography (Fig. 2). Themetallurgical approach involves the synthesis of alloys byrapid solidification that on subsequent annealing phasesegregates with a microstructure on the nanometer lengthscale. Such an approach is used for preparing very soft[8] commercial magnets or very hard [9] magnets. Theformer is achieved by creating nanoscale crystallites thatare randomly oriented but interact strongly to give a verysmall effective anisotropy by directional averaging. Onthe other hand, large anisotropies and coercivities veryclose to the theoretical maximum can be obtained by cre-ating isolated magnetic particles in a non-magnetic matrix[10, 11]. The alternative, well-known method to createnanoscale objects is by lithography. This top-down ap-proach, carried out with either electrons or photons, isboth time consuming and expensive but recent variationsalso include locally patterning a specific property, such asthe magnetic anisotropy, by implantation through a stencil

Figure 2 Alternative routes for the fabrication of zero-dimensional struc-tures. (a) Metallurgical—Hard Nd2Fe14B precipitates in a non-magneticNd matrix [10]; (b) Chemical—10 nm dia cobalt nanoparticles synthe-sized by chemical routes and self-assembled on a TEM carbon grid and (c)Lithography—arrays formed by ion-implantation through stencil masks—atop down approach [12].

mask [12]. The third approach is to chemically synthesizenanoparticles by the rapid decomposition of a metalor-ganic precursor by injecting it in a coordinating solventcontaining surfactants at elevated temperature [13, 14].Details of our work using the last method, including thesynthesis of core-shell particles, their magnetic behaviorand self-assembly, as well as their potential in biomedicalapplications are discussed later in this article.

The effect of dimensionality on magnetic behav-ior is best studied in thin film heterostructures. Their

794

40TH ANNIVERSARY

magnetic behavior is rich in new phenomena and includesrecent highlights such as surface/interface anisotropythat leads to perpendicular anisotropy in multilayerstructures [15] as well as frustration, proximity andinterface effects that lead to exchange bias (ferro-magnetic/antiferromagnetic bilayers) [16] and exchangespring (ferromagnetic/ferromagnetic bilayers) [17] be-havior. These magnetic properties are dominated by thephysical, chemical and magnetic structure of their inter-faces and it is important that appropriate tools be devel-oped for studying them most effectively at relevant lengthscales [18, 19].

The basic energies involved in the manipulation andcontrol of the magnetic properties of thin film het-erostructures are exchange (controls magnetic order) andanisotropy (controls magnetic orientation). These are phe-nomenological descriptions of fundamental correlationsand energies arising from the electronic and crystallinestructure of the material [20]. A soft ferromagnet (FM)such as iron has a large exchange parameter but a smallanisotropy, making ferromagnetic order stable at highertemperatures but with unpredictable orientation of mag-netization, especially if it is of nanoscale dimension. Onthe other hand, many antiferromagnets (AFM) have largeanisotropies and consequently, very stable orientations. InFM-AFM heterostructures, exchange coupling betweenthese layers can produce a ferromagnet with stable orderand high anisotropy, with the latter having unidirectionalcharacter—an unusual feature not observed in ferromag-nets. This phenomenon is called exchange bias (EB) be-cause the hysteresis loop is shifted and centered on anon-zero magnetic field. In spite of the simple physics,in reality EB films show a variety of unusual behaviorthat includes enhanced coercivity [21], magnetic trainingeffects [22], perpendicular coupling [23], kinked hystere-sis loops [24] and an asymmetry in the magnetizationreversal process [25]. A number of theoretical modelsproposed [26] have provided varying insight into the ex-change bias phenomenon but none of them have been ableto provide a comprehensive description of all its salientfeatures. Our recent work [27], involving epitaxial growthof high quality thin films, structural characterization byX-ray diffraction, neutron reflectivity and domain imag-ing using synchrotron radiation has shown that there isa fundamental asymmetry in the reversal mechanism ofsuch structures. On one branch of the hysteresis rever-sal takes place by domain nucleation whilst, in the otherdirection, it occurs by domain wall rotation.

For a ferromagnetic thin film or multilayer theanisotropy energy, Ean is written in terms of an effec-tive anisotropy constant, Keff as Ean ∼ Keff sin2θ . Keff

includes contributions from the demagnetizing field orshape anisotropy, surface/interface anisotropy and volumemagnetocrystalline anisotropy. For thick films (tM>fewnm for transition metals), the shape anisotropy energy

contribution is predominant and the magnetization liesin the plane of the film. For ultrathin films and mul-tilayers (tM<∼1 nm), it has been suggested that thesurface/interface anisotropy contribution, arising from abreak in symmetry at the interface [28] and proportionalto tM−1, can overcome the shape anisotropy and result ina spontaneous magnetization perpendicular to the film.Such perpendicular anisotropy in multilayers is now wellknown [15] but, in practice, an understanding of the ori-gin of this phenomenon requires a careful evaluation ofthe evolution and control of their complex microstruc-ture at the atomic scale [29]. Recently, we have combinedthese two phenomena, i.e. perpendicular anisotropy andexchange bias and critically evaluated the role of inter-faces in such perpendicularly exchange-biased systems.Details of our work in such thin film structures is alsodiscussed later.

Part of the interest in magnetic anisotropy and ex-change bias in thin films and multilayers is related tothe practically important area of spintronics, or spin elec-tronics [30]. Spintronics is an inter-disciplinary field ofresearch which includes the studies of electronic mate-rials/devices that explicitly employ the spin degree offreedom of the current carriers. Recent successful de-velopments include giant magnetoresistance (GMR) inmagnetic metal multilayers [31], spin valve devices [32],and tunneling magnetoresistance (TMR) in magnetic tun-nel junctions [33], that are intended primarily for appli-cations in magnetic field-sensing elements such as readheads and non-volatile magnetic random access mem-ories (MRAMs). Spin valve heads are already used inindustry; MRAM prototype structures have been demon-strated [34]. Spin manipulation in semiconductors is anattractive alternative for applications in the future gener-ations of sensors, memory, logic, and possibly quantumcomputing architectures. Many spintronic designs includeelectrical injection and sensing of magnetic moment as-sociated with the spin of the charge carrier. However,spin injection from a ferromagnetic metal electrode at-tached to a semiconductor structure is very inefficient dueto several reasons including simple impedance mismatchwhich is shown to lead to spin loss [35]. It has beenproposed that a ferromagnetic semiconductor with highly(ideally 100%) spin-polarized carriers can be effectivelyused for this application. Recent interest in ferromagneticsemiconductors, in particular the so-called “dilute mag-netic semiconductors” (DMS) [36, 37], is justified bothby their potential use as spin-injecting materials and thenew physics describing the properties of these materials.The best-studied DMS material so far, Mn-doped GaAs,was discovered in 1996, but is not likely to find practi-cal applications because of the low Curie temperature of∼170 K. Our fundamental studies are directed at under-standing the nature of ferromagnetism in wide band gapDMS with the Curie temperatures above room tempera-

795

40TH ANNIVERSARY

ture. In this article, we will briefly describe some of ourrecent work in this area. In particular, careful synthesisand detailed magnetic/transport measurements, comple-mented by advanced characterization of doped transitionmetal oxides has led to the discovery of a new class ofspintronics materials – “dilute magnetic dielectrics” orDMD [38]. These results also lead us to propose differentoperating principles for spintronic devices.

Finally, the optimization and understanding of the mag-netic and transport properties of this range of materials,be they nanoparticles or thin film heterostructures, re-quires a dynamic iteration between synthesis, propertymeasurements and the evaluation/control of microstruc-ture (physical, chemical & magnetic) at the appropriatelength scales [39]. There are three length scales that arerelevant: the characteristic length associated with the di-mensional characteristics of the magnetic phenomenonof interest, the size of the microstructural features aris-ing from the processing and the interaction length of theprobe being used to characterize these materials. It isimportant to identify the range where these three lengthscales overlap, for it is there that not only novel proper-ties and phenomenon are usually observed but a mean-ingful interpretation of the role of microstructure can bemade. For materials of interest in nanomagnetism and spinelectronics these length scales can range from the atomic(interfaces) to a few tens of nanometers (grain sizes).Hence, in addition to synthesis/processing, advances inthe development of such materials will critically dependon our ability to characterize them by a range of electron-optic and X-ray scattering/dichroism techniques. Theseinclude transmission electron microscopy [39], electronenergy loss spectroscopy with focused probes and the re-lated energy-filtered imaging methods [40, 41]; magneticspectroscopy [42] or imaging either in transmission [43]or photoemission [44] using synchrotron radiation andutilizing circular dichroism for magnetic contrast.

In the context of the above, we will present details ofour work in three distinct areas: magnetic nanoparticles,thin film heterostructures, and doped magnetic oxide thinfilms.

2. Magnetic nanoparticles and core-shellstructures

2.1. SynthesisMagnetic nanoparticles have drawn much attention re-cently due to their potential in magnetic recording aswell as many biological and medical applications suchas magnetic separation, hyperthermia treatment, mag-netic resonance contrast enhancement and drug delivery[45]. Synthesis of magnetic nanoparticles with controlledsize, shape, composition, morphology, and surface chem-istry is of great importance since their magnetic proper-ties are not only affected by all these factors, but they

also need to be specifically tuned depending on the re-quirements of a particular application. Using cobalt as amodel system [46, 47], we systematically studied the nu-cleation and growth of nanoparticles under various ther-modynamic/kinetic conditions and developed a generalmethodology to fabricate a variety of magnetic nanopar-ticles with desirable morphologies and controlled mag-netic properties. Cobalt nanocrystals are obtained by therapid thermal decomposition of cobalt carbonyl in a sur-factant mixture, forming the metal nanoparticle core andsoft organic shell. The synthesis initially follows the clas-sic La Mer mechanism [48] (short-burst nucleation andslow growth); however, further particle growth is stronglydependent on the available monomer concentration andthe existing surfactants in the system, which can resultin different morphologies of nanoparticles. The structureand properties of the surfactants play an important rolein the morphology, shape control—because the surfac-tants affect the reactivity of the monomers and the nucleiconcentration—and long term stability of the nanopar-ticles [49]. Our experimental results show that singlecrystal cobalt nanocrystals are obtained when oleic acid(OA)/trioctylphosphine oxide (TOPO) are used; however,using the oleic acid (OA)/dioctylamine (DOA) surfactantpair will give rise to multiple grain morphology. In addi-tion, disc-shape particles can be produced when oleic acid(OA) and oleylamine (ON) are present in the synthesis.

By adjusting the ratio of the concentration of surfac-tants to precursor and the relative bonding strength ofthe surfactants, we synthesize spherical cobalt nanocrys-tals with narrow size distribution in the size range of 5–25 nm, and anisotropic cobalt nanodiscs with specific as-pect ratios. Anisotropic cobalt nanoparticles are obtainedby rapid growth in a surfactant mixture under kinetic con-ditions, where the different surfactants are used to se-lectively control the growth rates of the different crystalfaces. Two typical examples of cobalt nanocrystals withdifferent morphologies are shown is Fig. 3. Here, cobaltnanodiscs appear like “nanorods” in TEM images, dueto the strong hydrophobic interaction between surfactanttails, which tends to draw them together. Therefore, diskswill stack face to face in order to maximize contact be-tween surfactant tails, and thus minimizing exposure toair. The magnetostatic energy is also a minimum whenthe neighboring disks lie face to face.

2.2. Self-assembly of cobalt nanocrystalsSelf-assembly is the spontaneous organization ofmolecules, molecular clusters or nanoscale objects into2D arrays and/or 3D networks by weak interactions. Formagnetic nanoparticles, in addition to the van der Waalsattraction between metallic cores and the repulsive forcesfrom the surfactant chains, magnetic interaction will alsoplay a critical role. Using cobalt nanocrystals as building

796

40TH ANNIVERSARY

Figure 3 Two examples of cobalt nanocrystals (A) 10 nm nanospheres and (B) 5×20 nm anisotropic nanodiscs.

Figure 4 Self-assembly behavior of Cobalt nanocrystals as a function of particle size, (A) square packing for 4 nm nanoparticles, (B) hexagonal closedpacking for 8–10 nm nanoparticles, (C) Bimodal distribution of two different size particles and (D) linear chains for 18 nm larger nanoparticles.

blocks, we systematically studied the self-assemblybehavior of these nanocrystals as a function of particlesize and shape. Depending on particle size and shape,one of the set of competing weak forces (steric, van derWaals, entropy and magnetostatic) will dominate theself-assembly and determine the resulting organization.In this fashion, we are able to selectively achieve squarepacking, hexagonal close packing, linear chains and for

bimodal size distributions of cobalt nanospheres spatialsegregation as a function of particle size [50] (Fig. 4). Asquare arrangement is observed for small size nanopar-ticles to minimize the repulsive steric forces betweenthe surfactant molecules on their surface. The interactionof surfactant chain significantly affects the organizationbehavior of these nanoparticles, since almost 50% of theatoms are on the surface and are bonded to the surfactants

797

40TH ANNIVERSARY

when particle size is smaller than 5 nm. As particlesize increases, the surface atoms become less and lessdominating. The nanoparticles behave like hard-spheresand self-assemble in a 2D hexagonal arrangementresulting from a first order phase transition as a functionof concentration. On the other hand, if these two sizesare combined to give a bimodal-size distribution thevibrational entropy of the system dominates. This leadsto a preferential wetting of the surface by the largerparticles such that the larger size particles are locatedin the center and are surrounded by the smaller sizenanoparticles. Such “fictitious” forces arising from theentropy of the system, called the depletion force, havebeen proposed in theoretical discussion of soft materials[51]. Cobalt nanocrystals become ferromagnetic whenparticle diameter is larger than 10 nm. Hence, as theparticle size increases, the magnetostatic interactionbecome significant, and linear chains or loops are formed.For anisotropic nanodiscs, lyotropic liquid-crystal-likearrays with increased orientation order as a function ofconcentration are observed in the TEM image (Fig. 5).

2.3. Core-shell nanocrystalsThe strategies developed for the synthesis of nanopar-ticles in homogenous solution can be extended to thegeneral case by separating the particle nucleation from itssubsequent growth stage. When the nuclei are differentfrom available monomer materials, particles composedof binary elements can be easily synthesized if conditionssuitable for heterogeneous nucleation can be generated insolution. The final morphology of such binary particles isdependent on their bulk thermodynamics: for immiscibleheterogeneous systems core-shell structures are obtainedwhilst miscible systems lead to alloy nanoparticles. Herewe are particularly interested in the core-shell structures,

since the biological applications of cobalt NPs are limitedby their poor biocompatibility and resistance to oxidation.It would be ideal if the shell could provide stability,biocompatibility and additional functionality. Most im-portantly, the shell materials should be immiscible withthe core materials. Considering all these requirements,gold becomes a natural choice for the shell material.

Gold coated magnetic nanoparticles have been reportedby different groups [52–54] However, in addition to repro-ducibility, the growth processes do not lend themselvesto the production of uniformly coated core-shell particlesbecause the synthesis environment is rich in oxygen andthe presence of water accelerates the formation of cobalthydroxide. Moreover, in this method, the use of a strongreducing agent (borohydride) makes the reduction reac-tion too rapid to form a uniform shell. It is also possiblethat instead of forming a shell, individual Au nanoparti-cles are formed.

Using pre-made cobalt nanoparticles as seeds, we growa gold shell in solution by slowly reducing a low-reactivitygold precursor with a weak reducer under mild conditions(85–105◦C) to form CocoreAushell nanoparticles. Subse-quently, these core-shell nanoparticles are made water-soluble by functionalizing them with hydrophilic thiol-containing surfactants, 11-mercaptoundecanoic acid,through the specific binding between gold and thiols.These CocoreAushell nanoparticles have been extensivelycharacterized by a wide range of transmission electronmicroscopy methods. The contrast of the particles in rou-tine transmission electron microscopy (TEM) images—lighter core and the darker shell—suggests the core-shellstructure is formed (Fig. 6A). High resolution TEM imageshow a single crystal Co core uniformly surrounded bymultiple gold grains, suggesting that gold has multiple nu-cleation sites on cobalt seeds during synthesis (Fig. 6B).The image clearly shows the structure of the shell, but the

Figure 5 Lyotropic liquid crystal behavior of cobalt naodiscs and show an increase in orientational order with increasing concentration (A, B, C).

798

40TH ANNIVERSARY

Figure 6 (A) 9 nm CocoreAushell nanoparticles: bright field TEM image, (B) high resolution TEM image, (C) EELS spectra of CocoreAushell nanoparticlesfrom center and (D) EELS spectra of CocoreAushell nanoparticles edge.

core is not well-resolved due to it being both in a differentcrystallographic orientation and embedded inside the thinshell. Lattice spacing of 0.204 nm and 0.102 nm, directlymeasured from the image, corresponds to fcc Au (002)and (004) planes. An inverse Fourier transform analysiswas performed on separate images from different regionsof the Au shell, and the projected symmetry of local im-ages fits well to fcc Au structure as well. The observationof Co-L2,3 edges in electron energy-loss spectroscopy us-ing a 1 nm probe focused on the core, (and absence ofthe peak when focused on the shell) confirms the distinctchemical nature of the core (Co) and shell (Au) (Fig. 6Cand D).

Complementing this detailed TEM analysis, bulk struc-tural and magnetic properties of these nanoparticles wereinvestigated on powder form samples by X-ray diffrac-tion (XRD) and superconducting quantum interferencedevice (SQUID) magnetometry, and the optical proper-ties in solution form by UV-Visible spectrophotometery.The relatively sharp peak of this core-shell structure in θ

− 2θ X-ray scans demonstrates a reasonably high degreeof crystallinity and uniform particle size of these core-shell nanoparticles (Fig. 7A). The temperature dependentmagnetization measurements show a narrow peak at 55 K,suggesting the magnetic size of the particle is about 6 nmwith a narrow size distribution by comparing with purecobalt nanoparticle measurement (Fig. 7B). At lower tem-peratures (5 K) the nanoparticles show hysteresis behaviorconsistent with the ferromagnetic state (Fig. 7B, inset).

In addition to the intrinsic magnetic properties of thecobalt core, the gold nanoshell brings in unique biocom-patibility and near infrared optical activity. The plasmon-derived optical resonance of the gold shell can be dra-matically shifted in wavelength from the visible regioninto infrared over a wavelength range that spans the re-gion of highest physiological transmissivity. From UV-visible spectra, pure Co nanoparticles show a continuousincrease in intensity with decreasing wavelength and nopeak is observed; pure Au nanoparticles show a charac-teristic peak around 530 nm; however, a relatively strongpeak at 680 nm was observed for the Au shell absorbanceof these Co-Au core-shell nanoparticles (Fig. 7C). Thisred shift from 530 nm to 680 nm compared to similar sizeindividual gold nanoparticles also suggests the existenceof the core-shell structures.

2.4. Biomedical applications of magneticnanoparticles: diagnostics andtherapeutics

The use of nanoparticles in biological applications is at-tractive because their small size allows them to interactwith biological molecules and systems in ways not pre-viously possible with larger objects. Nanoparticles madefrom magnetic materials have an added level of func-tionality. Their magnetic properties open up the poten-tial for novel treatments utilizing magnetic phenomena aswell as non-invasive and targeted therapy methods and

799

40TH ANNIVERSARY

Figure 7 CocoreAushell nanoparticles, (A) θ∼2θ X-ray scan. Inset is pure ε-Co spectrum, (B) ZFC/FC magnetic measurements, inset 5K hysteresis and (C)UV-visible spectra and the comparison with pure cobalt and gold nanoparticles.

Figure 8 Prior to use in bioapplications the surface of magnetic nanoparticles must be modified to provide functionality (specific binding is ideal). Afterlocalization at the target the magnetic properties of the particles provides novel functionality. Moving the particles with magnetic field gradients allows formagnetic targeting, separations and delivery. The particle’s magnetostatic field allows for diagnostics in vitro (microfluidics based biosensing) and in vivo(MRI contrast enhancement). The dynamic relaxation of the particles can be used for therapeutics (hyperthermia) or diagnostics (biosensing).

diagnostics (Fig. 8). For bioapplications there are threeparameters that will determine the performance of mag-netic nanoparticles: the surface chemistry, size (magneticcore, hydrodynamic volume and size distribution) and themagnetic properties of the particles. The surface chem-

istry and size will control the interaction of the particleswith biomolecules by providing biocompatibility, target-ing, controlled circulation time and specific binding. Thephysical and chemical properties of the particles them-selves, i.e. the size, size distribution, shape, crystallinity

800

40TH ANNIVERSARY

and composition, will control the magnetic properties andresponse of the particles to magnetic fields. Our work isrooted in our ability to control the morphology of theparticles as well as tailor their magnetic properties forspecific applications.

Movement of magnetic particles with a magnetic fieldgradient can be used to isolate, move or deliver biologicalmolecules or cells. Biomolecules of interest can be con-centrated in or separated from solutions via specific bind-ing to complementary biomolecules coated on the surfaceof magnetic nanoparticles. Drugs can be delivered locallywhen they are attached to the surfaces of the particles[55]. The particles can be concentrated at the target loca-tion either by specific binding to the target or by localizingthe particles magnetically. For this application magneticforces must be strong enough to overcome viscous drag.Currently, micron size clusters of superparamagnetic ironoxide particles, such as Dynabeads, are commerciallyavailable for separations. For applications which requiresmaller particles, nanoparticles must be made from mate-rials with high moments such as cobalt or iron-platinum[56]. The magnetic fields generated by the particles canbe used for diagnostics either in vivo or in vitro. In vivothe magnetic fields generated by the particles can affectthe relaxation of protons of surrounding molecules thusproviding contrast in MRI [57]. Companies such as Ad-vanced Magnetics have iron oxide nanoparticles commer-cially available with FDA approval for use as MRI contrastenhancers. In vitro magnetic nanoparticles can be boundto a surface via specific binding between biomoleculesof interest and their compliments attached to the surfaceof the nanoparticles. This binding event can be detectedwith a GMR [58] or planar Hall sensor [59] and used toidentify unknown biomolecules.

The relaxation of magnetic nanoparticles in an AC mag-netic field can be used for both diagnostics and therapeu-tics. Blocked particles relax via Brownian relaxation, witha relaxation time, τB given by τB = 3ηVH/kBT , whereη is the viscosity of the matrix and VH is the hydrody-namic volume of the particle (particle core plus surfacecoating and any associated layer). Brownian relaxationis dependent on the ability of the particles to rotate intheir matrix, thus changes in the viscosity [60 ] or thehydrodynamic radius [61] can be detected by measuringthe shift in frequency at which there is a peak in theimaginary component of the susceptibility. In this sensethe nanoparticles can be used as a biosensor to detectbinding events of biomolecules of interest on the sur-face of the functionalized particles. On the other hand,small unblocked particles relax via Neel relaxation, a ro-tation of the magnetic moment with an associated timeconstant, τN = √

π/2τ0 exp(K VM/kBT )/(K VM/kBT)1/2

which is dependent on the magnetic anisotropy of theparticles. Such relaxation of magnetic nanoparticles in anAC magnetic field can generate heat where the increase

in temperature is dependent on the time, frequency, theamplitude of the magnetic field, the relaxation dynamicsand the mechanism of thermal dissipation in the medium.For high heating rates the frequency must be correlatedto the relaxation and thus the properties of the magneticparticles. For bioapplications involving heat generation itis preferable to remain in the Neel relaxation regime as itis then possible to generate higher heating rates. Addition-ally, such heating is easier to model because the relaxationis independent of the surface coatings or mobility of thenanoparticles. Changes in mobility can occur if the parti-cles are located in tissue of varying viscosities (i.e. blood,muscle, fat) or if the particles are free or bound to a cellor protein.

In our work, in addition to metallic/alloy nanoparticlesand core-shell structures, we have also prepared highlyuniform, monodisperse, single crystal, magnetite Fe3O4,nanoparticles of tailorable size (4–11 nm) which we be-lieve are ideal for magnetic fluid hyperthermia. Hyper-thermia is the controlled heating of tissue to promotecell necrosis. The use of magnetic nanoparticles in hy-perthermia applications has been shown to be a powerfulcancer treatment [62, 63]. This method is non-invasiveand magnetic fields can be tuned to be invisible to tissue.We are developing methods to understand and predict thedynamic response and heating of the nanoparticles in anac-magnetic field [64]. After localizing the particles atthe target site, ac-magnetic fields will be applied. The fre-quency and amplitude of the field must be within biolog-ically compatible limits such that Ho f ≤ 4 × 108 A/ms[65]. The temperature of the target area will increase from37◦C to a target temperature of approximately 42–46◦Cfor approximately 30 min. Combining MRI imaging anddrug delivery potential of the magnetic nanoparticles withheating will result in a synergistic cancer treatment.

In order to generate the maximum heat for minimumdosage, it is critical to control the properties of thenanoparticle core and the surface chemistry. The ability ofmagnetic nanoparticles to generate heat is dependent onparticle size (magnetic core and hydrodynamic diameter),crystallinity, shape and particle polydispersity. Predictingthe dynamic behavior of ferrofluids in an ac-magnetic fieldis complex because of coupling between many parame-ters: the magnetic properties of individual nanoparticles,polydispersity of size and shape, ferrofluid concentration,temperature and viscosity of medium. Most real ferroflu-ids deviate from ideal behavior due to polydispersity ofsize and shape and particle interactions. Polydispersity ofparticle shape can have an effect on the anisotropy con-stants of the particles whereas particle interaction arisingfrom magnetostatic attraction or particle/solvent interac-tions can have a large effect on the hydrodynamic radiusof the particles. Consequently, polydispersed or agglom-erated samples can result in a mixture of relaxation behav-iors resulting in lowered heating rates [66] which may be

801

40TH ANNIVERSARY

Figure 9 Relative volumetric power dissipation as a function of particlesize calculated for our Fe3O4 nanoparticles at room temperature in water for10, 100, 500 and 1000 kHz. The peak near 11 nm is not highly dependent onfrequencies and indicates that 11 nm particles will have high heating ratesat therapeutic frequencies.

undesirable. Therefore, uniform, non-agglomerated andmonodisperse particles, such as ours, are ideal for mag-netic fluid hyperthermia. Preliminary models of the rela-tive volumetric power dissipation of our nanoparticles asa function of particle diameter indicate that peak heatingwill occur for particles around 11 nm for frequencies inthe 100 kHz–1 MHz range as shown in Fig. 9. This peakheating occurs for frequencies within a biologically com-patible range. By using a technique such as the slit-toroidmethod [67] the complex susceptibility of ferrofluids iscurrently being measured. Once the high frequency be-havior of our particles is quantified, we will model andstudy the change in temperature as a function of frequencyand field.

3. Thin film heterostructures: Exchange,proximity and interface effects

Advances in the past two decades in the synthesis andcharacterization of thin film heterostructures and sur-faces/interfaces has made it possible to prepare artifi-cially structured materials with tailored magnetic proper-ties [68]. Moreover, magnetism being a cooperative phe-nomenon, it can be readily manipulated in small structuresby the arrangement of different species with varying mag-netic interactions that includes proximity, exchange andinterface effects. Recent highlights include perpendicu-lar anisotropy observed in metallic multilayers comprisedof alternating ferromagnetic layers separated by a non-magnetic spacer and exchange bias or the observed shiftin the hysteresis loop of a ferromagnet grown in closeproximity to an antiferromagnet and field-cooled throughthe Neel temperature. Here we discuss some of our re-

cent work in this area spanning three broad topics: funda-mental mechanism of magnetization reversal in exchangebiased films, role of interface disorder in both perpendic-ular anisotropy and exchange bias and the nature of thecoupling between ultrathin ferromagnetic-ferromagneticbilayers.

3.1. Fundamental mechanism ofmagnetization reversals in exchangebiased films

Any systematic investigation of the phenomenon of ex-change bias in thin film form requires the selection of asuitable AFM and its successful growth in different crys-tallographic orientations and heterostructure configura-tions. We have identified the ordered intermetallic MnPdalloy with a L10 ordered structure and a tetragonal unit cell(a0∼0.407 nm, c0∼0.358 nm) as the appropriate AFM.For MnPd, the spin structure based on bulk measurements[69] is uncompensated (non-zero net spin on the interfaceplane, and with spin direction normal to the interface) forthe (100) surface but compensated (zero net spin on theinterface plane, but with spin direction also in plane) forthe (001) surface. We have successfully grown a MnPdfilm epitaxially on MgO (001) single crystals, using an Feprimary layer (t∼60 Å), with both a-axis and c-axis nor-mals [70]. Using alloy targets of similar compositions, wehave found that at low substrate temperatures the growthof MnPd is dominated by the kinetics and forms a dis-ordered structure, albeit with the a-axis normal. Thesefilms can be annealed to obtain a chemically orderedfilm with a (100) normal. At higher substrate tempera-tures, the growth process is thermodynamically drivenand forms the stable ordered structure with a c-axis nor-mal. On the other hand, growth of “normal” structures, i.e.MgO|MnPd|Fe can also be accomplished by growing atthe higher temperature range. The latter gives the largestexchange bias suggesting the use of texture as a predictiveprinciple for EB.

The growth and structural properties ofthe epitaxial film with orientation relation-ship given by MgO(001)||MnPd(001)||Fe(001) andMgO[100]||MnPd[100]||Fe[110] was investigated exten-sively by X-ray diffraction [71]. X-ray θ − 2θ scansshowed that the orientation of the MnPd unit cell andthe crystalline quality could be controlled as a functionof the deposition temperature. Pole figure measurementsor texture scans revealed a pure a-axis orientationfor films grown below 100◦C while single-crystallinec-axis films are obtained above 450◦C. Intermediatetemperatures yield a mixture of both orientations with apoor crystalline quality. Moreover, based on low angleX-ray reflectivity measurements, it was shown that theinterface quality strongly depends on the depositiontemperature and also the order in which MnPd and Fe

802

40TH ANNIVERSARY

Figure 10 Measured hysteresis loops with the applied field, H, along different crystallographic directions. (a) H applied along Fe [100] (bias direction), (b)H applied along Fe [110] (45 deg to the bias direction) and (c) along the Fe [010] direction (perpendicular to the bias direction).

are deposited. The X-ray reflectivity data (not shown)reveals finite thickness oscillations with two distinctwavelengths: the short wavelength oscillations are due tothe total thickness of the film while the longer ones arisefrom the Fe layer. Fitting the reflectivity data to a Parattmodel [72], indicates a MnPd/Fe interface roughness ofthe order of 2.5 Åfor our highest quality films grown witha substrate temperature above 450◦C.

The magnetization process of these high quality,exchange-biased MnPd/Fe bilayers was investigated us-ing vibrating sample and torque magnetometry. The for-mer was measured as a function of in-plane orientation,with respect to the directions of bias and easy axes ofmagnetization, and three distinct loops were observed(Fig. 10). A simple analytical model [73] based on co-herent magnetic moment rotation [74] was used to qual-itatively explain and describe the magnetization process.The shift of the hysteresis loop, the increased coercivity,the easy and hard axis behavior as well as the intermediatemagnetic state seen in the hysteresis loops are reproducedin the model. However, the magnitude of the bias and thecoercivity are not strictly in agreement with the measuredvalues. The discrepancies are attributed to the simplifiedmodel which does not take into account the role of mag-netic domains or disorder at the MnPd/Fe interface.

The magnetization processes in these exchange-biasedMnPd/Fe bilayers were further investigated using polar-ized neutron reflectivity [75]. The measurements showthat by breaking the symmetry of the intrinsic, cubic,four-fold anisotropy of the Fe film the induced unidi-rectional anisotropy radically changes the magnetizationprocesses. If the exchange bias is large the induced unidi-rectional anisotropy is able to pull the net magnetizationof the sample to the bias direction after saturation alongany of the magnetic hard Fe<110> directions. However,if the exchange bias is small it will only give rise to adifference in net magnetization along the magnetic easyaxes adjacent to the saturation direction. Thus, the netmagnetic moment along the bias direction depends on therelative magnitudes of the cubic and the unidirectionalanisotropy. However, both the magnetization and the neu-

tron reflectivity measurements show that the net magneticmoment along the bias direction at zero field is less thanthe total magnetic moment of the sample (M[100]<1).If the magnetization reversal occurs by coherent mag-netic moment rotation, M[100] would be equal to one.This suggests that magnetic domains are formed whenthe magnetic field is reduced from saturation. However,neither the neutron reflectivity nor the magnetization mea-surements provide any information on the distribution ofthe sample magnetization—only the components of thenet magnetization is measured—and a suitable imagingtechnique needs to be employed.

Photoemission electron microscopy (PEEM) was car-ried out utilizing the PEEM2 X-ray photoemission elec-tron microscope at the Advanced Light Source. In a pho-toemission electron microscope the sample is irradiatedby X-rays of variable polarization and photon energy.The resulting emission of low energy secondary electronsfrom the surface are then magnified onto a screen by elec-trostatic lenses and a spatial resolution [76] of typically50 nm can be achieved. Magnetic contrast of the Fe filmis obtained in such a microscope by using circular polar-ized X-rays and acquiring images at the Fe L3 (706.8eV)and Fe L2 (719.9 eV) core level absorption edges (X-raymagnetic circular dichroism or XMCD). Images shownhere are the result of a division of these two images. Be-cause the ratio between these two absorption intensitiesis determined by the projection of the magnetization ontothe helicity of the incoming light it allows for the un-ambiguous identification of the direction of the magneticmoment in a domain. The sample was saturated in a mag-netic field of 1500 Oe along a specific crystal axis beforethe magnetic domain images were taken.

PEEM images from the Fe/MnPd bilayer are shownin Fig. 11. It also shows a grey scale legend that linksthe direction of the magnetization to the image intensity.Note that the grey color corresponds to both horizontalmagnetization directions, since XMCD contrast appearsbetween “up” and “down” but not between “left” and“right” domains. The PEEM image for the descendingremnant magnetic state after saturation along Fe [110] is

803

40TH ANNIVERSARY

Figure 11 Photoemission electron microscopy images of exchange-biased samples at the remnant state on, (a) the descending and (b) the ascending branchof the hysteresis loops. The various crystallographic orientations and a guide to the direction of magnetization are also shown.

shown in Fig. 11A. No magnetic domains are observed inthe image. Thus, the unidirectional anisotropy (or EB) isstrong enough to pull the sample magnetization to the biasdirection (Fe [100]). For comparison, in an unbiased Fe(001) film the magnetization along the two magnetic easydirections next to Fe [110] is expected to be equal sincethe probability for the magnetization to rotate clockwiseand counterclockwise is the same.

The corresponding PEEM image for the ascending rem-nant magnetic state after saturation along the Fe [-1-10]direction is shown in Fig. 11B. Magnetic domains arein this case clearly visible. The magnetization in the do-mains is oriented along three of the four in-plane magneticeasy directions: Fe [0-10] (grey), Fe [-100] (black), and Fe[100] (white). As seen, most of the magnetization is alongthe Fe [0-10] direction, followed by the Fe [-100] and theFe [100]. The domains with the magnetization along thebias direction (Fe [100]) are due to the induced unidirec-tional anisotropy; they are not expected in an unbiasedFe (001) film after saturation along Fe [-1-10]. Thus, themagnetization along the bias direction is the reason for thelarge difference in remnance between the descending andascending loop. The domains with magnetization paral-lel and opposite to the bias direction are elongated with awidth of ∼0.5 µm and a length of ∼1–5 µm; moreover, thelong axes of these domains are perpendicular to each otherwith 90◦ Neel walls oriented along the Fe〈110〉 direc-tions. This type of domain wall is the most energeticallyfavorable type in ultrathin films with magnetic easy axesoriented 90◦ with respect to each other. Thus, the mag-netization is not along the long axis of the domains, butinstead makes an angle of 45◦ with it. It should be notedthat both the elongated shape and the perpendicular orien-tation of the Fe [-100] and the Fe [100] domains are due tothe fact that most of the sample magnetization is along theFe [0-10] direction. This means for the Fe [-100] domains

that only domain walls parallel to the Fe [1-10] (or equiv-alently Fe [-110]) direction are energetically favorable.Domain walls parallel to Fe [110] (or Fe [-1-10]) give riseto stray fields because the magnetization is not continuousacross these walls. Similarly, for the Fe [100] domains themagnetostatic energy is minimized if the domain walls areoriented along the Fe [110] (or Fe [-1-10]) direction.

Magnetic reversal in ferromagnetic materials occurseither by coherent rotation or by domain nucleation. Gen-erally, the mechanism is an intrinsic function of the mi-crostructure and is symmetric with respect to the appliedfield. In other words, on either branch of the hysteresisloop only one (and the same) of these two mechanisms isobserved. Based on these PEEM measurements we haveshown, for the first time, direct imaging evidence for theasymmetry in magnetic reversal mechanism in exchangebiased systems that is beyond the scope of indirect scat-tering measurements [77]. This was made possible by de-tailed X-ray photoemission electron microscopy imagingon carefully prepared samples which had been previouslywell-characterized magnetically including neutron reflec-tivity measurements. The magnetization reversal occursby moment rotation for decreasing fields while it pro-ceeds by domain nucleation and growth for increasingfields. The observed domains are consistent with the crys-tallography of the bilayers and favor a configuration thatminimizes the overall magnetostatic energy of the ferro-magnetic layer.

3.2. Exchange bias and perpendicularanisotropy in thin metal films andmultilayers

Combining the concepts of perpendicular anisotropy andexchange bias, we have investigated perpendicular ex-

804

40TH ANNIVERSARY

change bias, i.e. the coupling between Co/Pt multilay-ers with perpendicular anisotropy and antiferromagneticmaterials. Such investigations of perpendicular exchangebias would lead to a better understanding of the natureof the interfacial spin structure and magnetic coupling inFM/AFM structures. Initially a variety of critical growthparameters for both the AFM layer and the FM multilayerstacks, including the thicknesses of the Co and Pt lay-ers, seed layer materials, thin film orientations, number ofbilayer repeats and substrate temperature, were carefullyoptimized. As a result, high quality multilayers composedof Pt200 Å/(Co6 Å/Pt20 Å)5/FeMn were grown on Si (001)substrates at room temperature by ion-beam sputtering.An external magnetic field of 400 Oe, perpendicular tothe thin film plane, was applied during thin film growthusing a permanent magnet and a large perpendicular ex-change bias was observed.

In such thin film heterostructures, the physical prop-erties are sensitive to the structural parameters such asthe crystalline orientation, strain, roughness, and interdif-fusion at the interface. In particular, both the FM/AFMand the FM/NM interfaces are very important for perpen-dicular exchange biased multilayer systems and by vary-ing ion-beam energy, the interdiffusion at the FM/NMand FM/AFM interface can be well-controlled. The sam-ples were structurally characterized by X-ray reflectiv-ity (XRR) and in general, we found that the lower ion-beam energy deposited multilayers have lower interdif-fusion at the interface. Fig. 12 shows the XRR scans of(Co6 Å/Pt20 Å)5 multilayer samples deposited with differ-ent ion-beam energy of 250 eV, 500 eV and 750 eV. As theion-beam energy increases, the multilayers have a morerapid fall off of satellite intensity. The third-order satellitepeak can be clearly seen in Fig. 12A but is almost negli-gible in Fig. 12B and 12C. Associated data fitting usingthe commercially available Bede REFS (Build 4.00.13)program shows that the densities of the Co layers varyproportionally with ion-beam energy and, in all cases, are

Figure 12 X-ray reflectivity (XRR) of (Co/Pt)n multilayers grown withdifferent Ar+ ion-beam energies (a) 250 eV (b) 500 eV (c) 750 eV.

significantly higher than the bulk value. This is consis-tent with the fact that the energy of sputtered elementsincreases with increasing Ar+ ion energy leading to adeeper penetration and larger inter-diffusion.

The inter-diffusion effect on perpendicular exchangebias can be investigated by comparing the perpendicu-lar anisotropy energy and exchange-bias field of samplesgrown with different ion-beam energies. Fig. 13A showsthe room temperature, out-of-plane hysteresis loops of thesame three (Co6 Å/Pt20 Å)5 multilayers grown with ion-beam energy from 250 to 750 eV. As ion-beam energyincreases, the effective anisotropy constant change fromKeff = 4.17×106 erg/cm3 to Keff = 1.25×106 erg/cm3

and Keff = − 3.75×105 erg/cm3 for ion-beam energy of250 eV, 500 eV and 750 eV, respectively. From the mag-nitude and the change in sign of Keff , it can be seen thatthe transition from perpendicular anisotropy to in-planeoccurs with increasing ion-beam energy.

It is clear that the lower ion-beam energy enhances theperpendicular anisotropy with best results obtained for250 eV. The ion-beam energy also affects the exchange-bias field. Fig. 13B shows the hysteresis loops of repre-sentative multilayers of (Co6 Å/Pt20 Å)5/FeMn80 Å. All the(Co6 Å/Pt20 Å)5 stacks were grown with ion-beam energyof 250 eV, while the FeMn layers were grown with differ-ent ion-beam energies varying from 250 to 1500 eV. Theion-beam energy dependence of Hc and Heb is shown inthe inset of Fig. 13B. It is noticed that Heb drops rapidlyfrom 135 Oe to 45 Oe when the FeMn layers were grownwith higher ion-beam energy. Hc also decreases as the ion-beam energy increases, possibly due to the competitionbetween perpendicular anisotropy induced by the Co/Ptinterfaces and the large in-plane anisotropy induced at theterminating Co/FeMn interface together with the overallshape anisotropy of the film.

In conclusion, structural-property correlations havebeen investigated and it has been found that multilay-ers deposited with lower ion-beam energy have lowerinterdiffusion at the interface, and thus have stronger per-pendicular magnetic anisotropy and larger exchange bias.Work in progress include theoretical modeling of the ef-fect of interdiffusion on perpendicular exchange bias; ex-traordinary hall effect (EHE) [78] investigation of Co/Ptand (Co/Pt)/FeMn multilayers to study the sensitivity ofelectron transport to the magnetic multilayers interfacesand magnetic force microscopy (MFM) investigation ofthe domain formation and motion during the magnetiza-tion reversal process.

3.3. Perpendicular domain formationand transport characteristicsof ferromagnetically coupled bilayers

In magnetic thin films, the orientations of the magneticmoments can be perpendicular or parallel to the surface of

805

40TH ANNIVERSARY

Figure 13 Out-of-plane hysteresis loops of (a) (Co/Pt)n multilayers without bias grown with different Ar+ ion-beam energies. (b) (Co/Pt)n/FeMn multilayersin which FeMn was deposited with different ion-beam energies. Inset, exchange bias field (Heb) and coercivity (Hc) versus ion-beam energy.

the system and is commonly called out-of-plane and in-plane, respectively. Most ferromagnetic metal thin filmsshow in-plane domain structures [79] unless their filmthickness is less than 1∼2 nm. However, a stripe domainstructure [80] with significant out-of-plane character canbe achieved in thin film form, when materials such asYttrium Iron Garnet (YIG) are grown along certain crys-tallographic direction because of its high perpendicularmagnetocrystalline anisotropy. If a ferromagnetic film isthen grown on top of such a garnet film exhibiting stripedomains, up to a certain thickness (<20 nm), it is expectedto mimic the domain structure of the garnet underlayer.In our investigation, the perpendicular domain couplingbetween a ferromagnetic metal and the garnet underlayerhas been studied in detail. Micromagnetic simulationsshow that the exchange stiffness energy term, which isthe interaction energy between nearest-neighbor atoms atthe interface, is the main cause of a perpendicular domainstructure in Fe films rather than a long range surface ex-change energy term. Finally, work in progress includesmeasurements of the magnetoresistance including the do-main wall resistance due to local spin-dependent elec-tron scattering at the domain wall. In principle, using thissystem the position of the domain wall in the metallicferromagnet can be modulated by the underlying stripedomain of the garnet layer but the domain wall-width isdetermined by the intrinsic properties of the metallic fer-romagnet. Since the domain wall of this ferromagneticmetal layer is perpendicular to the current direction, themagnetoresistance due to domain wall scattering can bemeasured.

3.3.1. Experimental detailsIn this preliminary experiment, a commercial YIG layergrown by liquid phase epitaxy (LPE) on a gadolinium gal-

lium garnet (GGG) single crystal was used as substrate. Tostudy the domain coupling between ferromagnetic metaland magnetic oxide layers, a series of test samples wereprepared using an ion beam sputtering system. A softferromagnetic metal Fe was deposited on the YIG/GGGsubstrate at room temperature. The thickness of Fe, tFe wasvaried (0, 5, 10 and 20 nm). Magnetic hysteresis of suchthin films were measured (Fig. 14) using the magneto-optic Kerr effect (MOKE) [81]. Depending on the direc-tion of magnetization and the direction of light propaga-tion, both in-plane and out-of-plane hysteresis loops weremeasured. A square-shaped hysteresis loop for out-of-plane measurements was observed for all tFe<20 nm. FortFe=20 nm, the out-of-plane loop shows a linear shape,but the in-plane loop shows a square-shaped hysteresis,indicating that the preferred magnetization direction isin-plane.

Magnetic Force Microscopy (MFM) is a powerful tech-nique to investigate the surface domain structure of mag-netic materials. MFM uses a two-pass technique for imag-ing the domain structure. The first pass measures the to-pography using contact or tapping mode, and the secondpass measures the interaction force between tip and sam-ple using lift mode. Thus, MFM allows a detailed corre-lation of magnetic and structural features at the surface.Both topography (Atomic Force Microscopy or AFM)and MFM images of the Fe/YIG samples are shown inFig. 15. In the AFM images, as the thickness of Fe layeris increased the surface of the film appears to be gettingsmoother. The corresponding MFM images (Fig. 15) re-veal a stripe domain structures for tFe = 5, 10 nm, butnot for tFe = 20 nm. Since the stripe domain imagesare the characteristics of magnetization structures withsignificant out-of-plane contribution, these results are ingood agreement with the results from the MOKE mea-surements. Also, the domain patterns of both 5 nm and10 nm Fe films are very similar to that of the YIG only

806

40TH ANNIVERSARY

Figure 14 MOKE results of Fe/YIG/GGG sample with Fe thickness of 0 nm, 5 nm, 10 nm, 20 nm. The hysteresis loops are measured for both (a)out-of-plane and (b) in-plane.

Figure 15 Topography (AFM) and domain structure (MFM) of Fe/YIG/GGG samples with Fe thickness of (a) 0 nm, (b) 5 nm, (c) 10 nm, (d) 20 nm.(Scanning area: 40 µm × 40 µm).

sample in both shape and dimension. This suggests thatthe domain structure of YIG film is maintained in theFe film due to the strong magnetic coupling between thelayers [82].

3.3.2. Micromagnetic simulationThe experimental results show that the domain configura-tion observed in Fe/YIG bilayer structures arises from astrong interlayer exchange interaction between the Fe andYIG layers. To verify these results, micromagnetic simu-lation was performed on Fe 10 nm/YIG double layer con-figuration using OOMMF (Object Oriented Micromag-netic Framework) [83]. The volume exchange stiffnessenergy and the interface energy term with both bilinear

and biquadratic contributions are the two energy terms,which determine the domain structure of Fe/YIG sam-ple. Bilinear exchange coupling is the interlayer coupling,which leads to antiparallel alignment of the magnetic lay-ers, and biquadratic exchange coupling is the interlayercoupling, which leads to a 90◦ relative orientation of themagnetization in adjacent magnetic layers. The volumeexchange stiffness energy is expressed as [84]:

Eex = Ainterface

∫(∇ · m)2dV (1)

where Ainterface, the exchange stiffness constant (J/m) atthe interface. m is the normalized unit spins (i.e. magne-tization direction) at each cell. Another energy term, the

807

40TH ANNIVERSARY

interface contribution to the energy density between twothin films can be expressed as [85 , 86]:

Ecoupl = Cbl (1 − m1 · m2) + Cbq[1 − (m1 · m2)2] (2)

where Cbl and Cbq are the bilinear and biquadratic sur-face exchange constants, and m1 and m2 are the normal-ized, unit spins (i.e., magnetization directions) at adja-cent cells in two different materials. The Fe/YIG bilayerdomain structure was simulated while varying the val-ues of Ainterface, Cbl and Cbq and keeping all other ma-terials parameters fixed to their bulk values. For fixedAinterface, but with the bilinear and biquadratic surface ex-change constants varied over a wide range, from 0 to1×10−4 J/m2, the simulations always show an in planedomain configuration for the Fe film. On the other hand,a strong interlayer coupling between the Fe and YIG film,which shows the perpendicular domain configuration ofthe Fe film when the interface exchange stiffness coeffi-cient (Ainterface) is varied. These simulations show that theobserved interlayer coupling between Fe and YIG layerscan be attributed to a strong volume exchange stiffnessenergy contribution.

4. Wide band-gap, doped, magnetic oxides forspintronics applications

The rapidly developing field of spin electronics requires asemiconducting room temperature ferromagnet for incor-poration in proposed spin-electronic devices. There arecertain materials requirements in order to have a spin-tronic device. These are: efficient electrical injection ofspin-polarized carriers (spin injection) into the semicon-ductor, sufficient spin diffusion lengths and lifetimes forspin transport through the semiconductor, control and ma-nipulation of the spin signal, and efficient spin detec-tion for determining output [87]. In addition, for any sortof commercial device, room temperature operation is amust. Materials that have the potential to satisfy these re-quirements are dilute magnetic semiconductors (DMS).A DMS is composed of a nonmagnetic semiconduct-ing matrix with spins of the magnetic (transition metal)dopant ions coupled ferromagnetically, and ideally, pro-viding spin polarized charge carriers for conduction of thesignal. In this context, studies of transition-metal doped,wide band-gap materials are of current interest both in thesolid-state physics and materials science communities.Our work in this field is both active and varied. Initialwork was focused on multilayers comprised of alternat-ing nanoscale layers of a semiconductor and a magneticmetal. This resulted in samples ranging from a superpara-magnetic granular material to a DMS multilayer stack.Subsequently, we have progressed to the incorporation ofa randomly distributed magnetic dopant within a semi-

conducting matrix resulting in dilute magnetic dielectrics(DMD) [38].

DMS such as Mn:GaAs have received the most atten-tion as spin injector materials for spintronics [36, 37]because they allow for impedance matching to the semi-conductor device, a requirement for efficient injection[35]. The functionality of such devices has been limitedthus far by low ferromagnetic ordering temperatures, withTC∼170 K. In the search for materials compatible withroom temperature operation, there has been significantfocus recently on wide band gap semiconductors such asTiO2 [88, 89], ZnO [90], GaN [91, 92], GaP [93], AlN[94, 95], and SnO2 [96, 97] as the host DMS material.The two most prominent issues that limit further develop-ment of such transition metal (TM) doped oxides for roomtemperature spintronics are difficulties that can arise fromphase segregation and lack of understanding of the fun-damental mechanism associated with their ferromagneticordering. Phase segregation of transition metal dopants(which can act as extrinsic sources of ferromagnetism) inthe oxide host has to be strictly avoided. Instead, prepa-ration of a homogeneous solid solution of the TM dopantincorporated into the oxide host is necessary to determineif these materials do in fact exhibit intrinsic ferromag-netic behavior. It is important to note that in the case ofsecondary phase formation of the magnetic dopant atoms,insufficient characterization leads to skewed results, caus-ing much controversy in determining whether these com-plex material systems can be used as true DMS. Even inthe cases where secondary phase formation can be ruledout, there has been a wide range of results for the samematerial grown and processed under different conditionsand in different laboratories, which makes it difficult topropose a comprehensive explanation for their high TC

ferromagnetic behavior. These issues continue to presentthemselves as an exciting topic in the field of magneticmaterials, and much still needs to be investigated bothexperimentally and theoretically.

4.1. Co doped anatase TiO2—a dilutemagnetic dielectric

With the initial discovery of room temperature ferromag-netism in Co doped anatase TiO2 [88, 89], this compoundhas been one of the most prominent of the potential oxideDMS materials [98]. However, due to the observation ofCo metal clusters in some cases, there has been much spec-ulation on the origin of ferromagnetism in such TM dopedoxides. On the other hand, some recent experiments havealso ruled out the presence of clusters, confirmed roomtemperature ferromagnetism in these materials [92, 93]and correlated the ferromagnetism to n-type semiconduct-ing behavior. Although these results appear promising forCo:TiO2 as a potential DMS, the true origin of ferromag-netism and whether the carriers are spin-polarized are yet

808

40TH ANNIVERSARY

to be determined. It is hence of vital importance to clarifythe correlation (if any) between carriers and the mech-anism of ferromagnetism inherent to this structure. To-wards this end, thin films of Co-doped anatase were grownby RF magnetron sputter deposition. Because anatase isa metastable phase of TiO2, films need to be grown atelevated temperatures (550◦C) at relatively slow growthrates (0.01 nm/s) and on lattice-matched substrates. Herewe present results for films grown on LaAlO3 (100) orLAO. The films were then annealed in ultra high vacuum(UHV) for 1 h at 450◦C. Structural, compositional, andmagnetic measurements were made on films before andafter annealing.

X-ray diffraction (XRD) verified that the as-depositedfilms were single phase anatase with no evidence of sec-ondary phases within the detection limits of X-ray scatter-ing experiments. The effect of the post-growth annealingtreatment in UHV at 450◦C for 1 h is shown by the changein the anatase (004) peak from the XRD θ−2θ scan inFig. 16. The FWHM value of the film peak is shown to de-crease by −0.20◦ after annealing, which illustrates the im-

Figure 16 XRD 2θ scan of Co:TiO2 anatase (004) peak before and afterannealing.

provement in crystalline quality of the film. Also, it can beseen that the peak position shifts to the right with anneal-ing, indicative of a relaxation of the anatase lattice alongthe c-direction. This is most likely attributable to strainrelaxation, diffusion of defects through the lattice (oxygenvacancies) and/or incorporation of Co ions into equi-librium lattice positions. Fig. 17 demonstrates the highquality anatase crystal structure observed in high resolu-tion transmission electron microscopy (HRTEM) imagesof cross-sectional as-deposited (A) and UHV annealed(B) specimens. The selected area diffraction pattern (in-set) reveals single crystal (004) anatase epitaxially grownonto the LAO substrate. EDXS data taken at a number oflocations throughout the specimens reveal a solid solutionof Co dissolved in anatase, with Co concentrations rang-ing from ∼2–7 at.% incorporated into the lattice. Analysisof energy filtered images using hydrogenic cross-sectionsalso shows a uniform distribution of cobalt with an aver-age concentration of 2.8% Co in the anatase lattice [99].Rutherford backscattering spectroscopy measurementsconfirm that the bulk Co composition as a function offilm depth is uniform with an average of ∼2 at% [100].Moreover, there was no evidence for the presence of Cometal or Co-rich (above 7 at%) anatase clusters withinthe films in any of these detailed TEM measurements.

Near edge X-ray absorption fine structure (NEXAFS)at the Co K-edge were measured on annealed and as-deposited Co:TiO2 films to determine the oxidation stateand local geometry of the Co dopant in the lattice. As-deposited and annealed films exhibited nearly identicalNEXAFS spectra; for clarity Fig. 18 [38] shows only theannealed sample spectrum compared to the following ref-erence samples: CoTiO3, CoO, and Co metal. The onset ofabsorption in the NEXAFS spectra of the Co:TiO2 filmsare a close match to CoTiO3 and CoO, where Co has beeninterpreted to be in the +2 oxidation state. On the other

Figure 17 High resolution TEM images from cross-section specimens of as-deposited (A) and UHV annealed (B) Co:TiO2 films on LAO. Insets: Lowmagnification images, and selected area diffraction pattern.

809

40TH ANNIVERSARY

Figure 18 Co K-edge NEXAFS spectra for UHV annealed Co:TiO2 film with reference samples: Co:TiO2 film (open circles), CoTiO3 (solid line), CoO(long dash), and Co metal (short dash) spectra (reproduced from [38]).

hand, the absorption edge of Co metal is a poor match tothe Co:TiO2 films due to the much lower threshold en-ergy corresponding to the Co(0) state. The closer matchof NEXAFS spectrum of the sputtered Co:TiO2 film toCoTiO3 than to CoO is indicative of the distorted octahe-dral coordination of the Co atom in the lattice (i.e. Ti sitein anatase), whereas in CoO the Co site has an undistortedoctahedral configuration. From this comparison of Co K-edge spectra it can be concluded that there is no evidenceof Co metal throughout the film and Co2+ substitutes forTi4+ in the lattice.

Magnetic hysteresis loops (M vs H) at 300 K of as-deposited and annealed Co:TiO2 films in Fig. 19A showthe increase in spontaneous magnetization, MS, from 0.24to 1.17 µB/Co atom with annealing. The TRM measure-ments (Fig. 19B) were taken in zero applied field afterthe sample was saturated, and the remanent magnetiza-tion, MR, as a function of increasing temperature wasrecorded. (TRM is useful as a direct measurement of themagnetic signal from the Co:TiO2 film alone because themeasurement is taken in zero applied field and there is nocontribution from the diamagnetic LAO substrate.) TheTRM results from 5–365 K demonstrate the enhancementof MR from 0.07 to 0.34 µB/Co atom with annealing. Co-ercivity, HC, also increases with annealing, from ∼100 to∼300 Oe.

Both as-deposited and annealed samples are all highlyinsulating and have sheet resistances ≥1011 �/square.Thus resistivity is at least greater than 106 � cm forall samples. The fascinating co-existence of room tem-perature ferromagnetism in the dielectric state brings tolight many questions regarding the possible ferromagnetic

exchange mechanisms in TM doped oxides. The XRD ex-perimental results suggest that the improved crystallinityof the films after UHV annealing would be the primaryfactor for the enhanced ferromagnetism. However, the450◦C annealing process in UHV also provides thermalenergy for the creation and diffusion of oxygen vacancies(VO’s) through the lattice. The VO‘s created during the an-nealing that are necessary for the enhanced ferromagneticproperties do not provide free carriers for conduction.Rather, we propose that the carriers associated with cre-ation of Vo’s are trapped on CoTi defect sites [101, 102].With sufficient thermal energy provided by the annealingtreatment, both the VO migration and trapping is possi-ble, leading to the enhancement of magnetic properties.We therefore attribute the enhanced ferromagnetism tothe combination of the increase in crystallinity, the cre-ation/diffusion of defect centers and trapping of carrierswith the annealing process.

4.2. Other examples of DMD - Co:ZnOand Cr:ZnO systems

Alternating deposition of atomic-scale layers of ZnO andCo, with aluminum doping, with varying semiconduc-tor and metal thicknesses results in a range of magneticand electronic properties [104]. A series of multilayersamples were deposited via Ion Beam Sputtering (IBS),ZnO thicknesses ranged from 2–20 Å and Co thicknessesranged from 2–10 Å with the total number of bilayersvarying between 15 and 25. The samples were depositedat constant deposition rates on Si and glass substrates.Additionally, through collaboration, we were able to com-

810

40TH ANNIVERSARY

Figure 19 Magnetic measurements from SQUID magnetometer: (A) M-H loops at 300 K of as-deposited (dashed line) and UHV annealed (solid line)Co:TiO2 films. (The diamagnetic contribution from the LAO substrate has been subtracted from the experimental data.) (B) MR vs. T of same films.

pare this multilayer system with equivalent single crystalZnO:Co samples prepared by the metalorganic chemicalvapor deposition (MOCVD) process [105].

RBS studies showed the Co/Zn atomic ratios rangingfrom 0.1 to 1.25, consistent with the expected valuesdepending on the details of the multilayer stack, anexcess of oxygen compared to Zn in stoichiometricZnO, and about 5% of aluminum (donor dopant). XRDfor a (ZnO

20 ÅCo2 Å)25 sample on a glass substrate

reveals the nanocrystalline structure of ZnO, primarilyc-axis oriented, and exhibiting no peaks of metallic Co.Samples with larger relative Co content indicated diffusemetallic Co reflections. Zero field-cooled and field-cooledmagnetization measurements were made as a function oftemperature in a field of 20 Oe (Fig. 20) for samples withfixed nominal thickness of 2 Å of Co layers and varying(2–20 Å) ZnO thickness with magnetization normalizedto the nominal total volume of Co. Higher relativecontent of Co results in a granular multilayer structure,which demonstrates magnetic blocking phenomena andsuperparamagnetism. Blocking is seen from splittingof the zero field cooled (ZFC) and field cooled (FC)curves, and, for the sample with the thinnest ZnO layers,(ZnO2 ÅCo2 Å)25, a maximum in the ZFC susceptibilitynear the blocking temperature. The sample with thethickest ZnO layers, (ZnO20 ÅCo2 Å)25, behaves as adiluted magnetic semiconductor, i.e. it does not showany sign of blocking phenomena down to 2.5 K, and itsmagnetization does not go to zero at high T, and remainsapproximately constant in the temperature range 40–300 K. M vs H dependences for this sample are hystereticboth at 10 K and 300 K. Loops with no hysteresis typicalof superparamagnetic materials are observed above theblocking temperature in samples with small relative ZnOthickness (2–10 Å) to Co (2 Å). The transition from gran-ular 2D to magnetic semiconductor 3D material is furtherevidenced by the following transport measurements.

Resistivity of granular samples with high relative Coconcentrations follows a variable range hopping (VRH)

Figure 20 ZFC and FC magnetization measurements at H=20 Oe forsamples with varying nominal thickness of ZnO layers x (ZnO

xACo2A)25,

normalized to the total nominal volume of Co metal. Inset: Magnetic hys-teresis loops measured on sample (ZnO20 ACo2 A)25 at 10 K and 300 K.

law (ln ρ∼(T0/T)n+constant) with n close to 1/2, typ-ical of hopping between metal granules in an insulatorat intermediate temperatures [105] evidenced in Fig. 21.We’ve also shown [104] that a crossover from the dom-inating quasi-2D (n>1/2) to 3D (n<1/2) variable rangehopping occurs as the ZnO nominal thickness increases.Upon further increase of ZnO thickness, with n∼1, theferromagnetic sample (ZnO20 ÅCo2 Å)25 behaves like thatof an Al doped ZnO film with no Co doping.

There are two competing channels of conduction in thesystem, which differ in the temperature dependence of re-sistivity. Increasing the total relative amount of Co in thefilms, at a fixed Al doping level, increases the resistivity ofthe semiconductor. Hence, cobalt impurities, while neces-sary for the presence of magnetic moments, provide trapsfor carriers which mediate magnetic order. With furtherdecreasing ZnO layer thickness, the number of Co metal

811

40TH ANNIVERSARY

Figure 21 Logarithm of resistivity as a function of 1/T1/2.

inclusions grows, which leads to increasing conductivityvia inter-granular tunneling (hopping) channel. The fer-romagnetic sample (ZnO20 ÅCo2 Å)25 contains no metalinclusions, based on both magnetic and structural results.We assume that, due to limited diffusion length at a givenambient deposition conditions, the semiconductor super-lattice consists of cobalt-rich, low carrier-concentrationlayers, alternating with layers lightly doped with Co,and hence having higher carrier concentration. Thoughconducting due to Al doping, neither ZnO:Co bulk filmsgrown by MOCVD, or the superlattice films already cov-ered, show an appreciable anomalous Hall contribution.Additionally, magnetically doped ZnO DMS films showqualitatively similar magnetoresistance to undoped films[104]. This suggests that the ferromagnetism in DMS

Figure 22 Hysteresis loops of as grown and annealed samples showingincrease in Ms, MR. The inset shows the coercive fields.

films is not carrier mediated, indicating that a differentferromagnetic exchange mechanism is at play.

Magnetic doping of wide gap semiconductors includingZnO and TiO2 bearing no relationship to the presence offree carriers has resulted in a new class of spintronic mate-rials tentatively titled dilute magnetic dielectrics (DMD)as discussed above. As an additional proof of this concept,thin films of ZnO doped with Cr grown via magnetronco-sputtering have resulted in magnetic dielectric films[106]. Grown on a-plane sapphire, c-axis oriented poly-crystalline ZnO films with Cr concentration of roughly9.5 at.% show a per-ion saturation moment of 1.4 µB. Thefilms show open M-H plots (Fig. 22) and remanent mag-netization at temperatures exceeding 365 K. The lack ofconductivity excludes the possibility of carrier-mediatedferromagnetism, and structural characterization (XRD)does not indicate the existence of any secondary phases(to contribute to the ferromagnetism in these films, a sec-ondary phase would have to make up a significant volumefraction of the film).

5. Concluding remarksResearch in the area of magnetism and magnetic mate-rials has undergone a tremendous transformation duringthe last couple of decades. It is a rich combination of newtheoretical concepts, synthesis, measurement and charac-terization techniques, where both the materials structureand “device” engineering is carried out at the nanometerlength scales. Magnetic materials are also central to newapplications ranging broadly from information storage,biology and medicine to new principles of electronics andeven quantum computing. In addition to traditional stud-ies of magnetic soft/hard materials, current scientific in-terest is also focused on magnetic quantum dots, superlat-tices, precision doping and magneto-electronics. Hence,many quantum phenomena more familiar to semiconduc-tor physics also play a vital role in these magnetic stud-ies. This also means that the whole area of magnetismis becoming more synthetic, multidisciplinary and intel-lectually demanding. In this brief overview of our recentwork we have demonstrated broadly this blend of ma-terials synthesis, microstructure, new physics and novelapplications. However, the format of this article does notallow the discussion of some other interesting work beingcarried out in our laboratory. This includes, for exam-ple, studies of materials for magnetic actuations in mi-croelectromechanical systems (MEMS) applications, andspin-resolved quantum conductance in nanoscale breakjunctions. Details of these projects and our broad researchprogram can be found on our website [107].

AcknowledgementsOur research program at the University of Washington wassupported by the following grants: NSF/DMR #0203069,

812

40TH ANNIVERSARY

NSF/DMR #0322729, NSF/ECS #0224138, DoE/BES#DE-FG03-02ER45987, NSF/DMR #0315460. KMKalso acknowledges the Campbell endowment at UW fortheir continued support. Graduate students involved inthis research were also supported by the UW-PNNL JINfellowship (KG, YB) and the NPSC/ARCS/Boeing Engi-neering Scholarship (MG). We thank all our collaboratorsfor their help and support

References1. G . H A D J I PA NAY I S (ed.), Magnetic Recording Beyond 2000,

(Kluwer Academic Press, 2001).2. L . M. FA L I C OV, D. T. P I E R C E, S . D . BA D E R, R .

G RO N S K Y, K. B . H AT H AWAY, H. J . H O P S T E R, D. N.L A M B E T H, S . S . P. PA R K I N, G. P R I N Z, M. S A L A M O N,I . K . S C H U L L E R and R. H. V I C TO R A , J. Mater. Res. 5 (1990)1299.

3. F. J . H I M P S E L, J . E . O RT E G A, G. J . M A N K E Y and R.F. W I L L I S , Adv. Phys. 47 (1998) 511.

4. H . K RO N M U L L E R in Modern Problems in Metal Physics II”edited by A. Seeger (Springer, Berlin, 1966).

5. C . K I T T E L , Rev. Mod. Phys. 21 (1949) 541.6. E . H . F R E I , S . S H T R I K M A N and D. T R E V E S , Phys Rev.

106 (1957) 446.7. C . P. B E A N , J. Appl. Phys. 26 (1955) 1381.8. G . H E RT Z E R , IEEE Trans. Mag. 25 (1989) 3327.9. J . M. D. C O E Y (ed.), Rare-Earth Iron Permanent Magnets

(Oxford, 1996).10. D . C . C R E W, E R. G I RT, M. G U I L OT, D. S U E S S , T.

S C H R E F L and K. M. K R I S H NA N , Phys. Rev. B66 (2002)184418.

11. E . G I RT, K. M. K R I S H NA N and G. T H O M A S , ScriptaMaterialia 44 (2001) 1431.

12. G . J . K U S I N S K I , K . M. K R I S H NA N, G. D E N B E AU X,G. T H O M A S, B . D. T E R R I S and D. W E L L E R , Appl. Phys.Lett. 79 (2001) 2211.

13. V. K . L A M E R and R. H. D I N E G A R , J. Amer. Chem. Soc. 72(1950) 4847.

14. L . B RU S , J. Phys. Chem. Solids 59 (1998) 459.15. H . J . G . D R A A I S M A, W. J . M. D E J O N G H E and F. J . A .

D E N B RO E D E R , J. Mag. Mag. Mater. 66 (1987) 351.16. J . N O G U E S and I . K . S C H U L L E R , ibid. 192 (1999) 203.17. E . F. K N E L L E R and R. H AW I G , IEEE Trans. Mag. 27 (1991)

3588.18. J . B . KO RT R I G H T, D. D. AW S H A L O M, J . S TO H R, S .

D . BA D E R, Y. U. I D Z E R DA, S . S . P. PA R K I N, I . K .S C H U L L E R and H. C . S I E G M A N , J. Mag. Mag. Mat. 20 (1999)7.

19. K . M. K R I S H NA N , in NATO/ASI on Magnetic Recording Beyond2000, edited by G. Hadjipanayis (Kluwer Academic Press, 2001) p.251.

20. A . M O R R I S H , The Physical Principles of Magnetism (Wiley, NY,1965).

21. C . L E I G H TO N, J . N O G U E S, B . J . J O N S S O N- A K E R M A N

and I VA N K. S C H U L L E R , Phys. Rev. Lett. 84 (2000) 3466.22. U . W E L P, S . G . E . T E V E LT H U I S , G . P. F E L C H E R, T.

G R E D I G and E . D. DA H L B E R G , J. App. Phys. 93 (2003) 7726;S . G . E . T E V E LT H U I S , A . B E R G E R, G. P. F E L C H E R,B . K. H I L L and E . DA N DA H L B E R G , ibid. 87 (2000) 5046.

23. Y. I J I R I , J . A . B O R C H E R S, R . W. E RW I N, S . H . L E E ,P. J . VA N D E R Z A AG and R. M. W O L F , Phys. Rev. Lett. 80(1998) 608.

24. P. B L O M Q V I S T, K A N NA N M. K R I S H NA N and E R.

G I RT , J. Appl. Phys. 95 (2004) 8487; C . L E I G H TO N, M.R. F I T Z S I M M O N S, P. YA S H A R, A. H O F F M A N, J .N O G U E S, J . D U R A, C . F. M A J K R Z A K and I VA N K.S C H U L L E R , Phys. Rev. Lett. 86 (2001) 4394.

25. M. R . F I T Z S I M M O N S, P. YA S H A R, C . L E I G H TO N,I VA N K. S C H U L L E R, J . N O G U E S, C . F. M A J K R Z A K

and J . A . D U R A , ibid. 84 (2000) 3986; M. G I E R L I N G S, M.J . P R A N D O L I N I , H . F R I T Z S C H E, M. G RU Y T E R S, D.R I E G E L , Phys. Rev. B65, (2002) 092407; V. I . N I K I T E N KO, V.S . G O R NA KOV, A. J . S H A P I RO, R . D. S H U L L, K. L I U ,S . M. Z H O U and C. L . C H I E N , Phys. Rev. Lett. 84 (2000) 765;V. I . N I K I T E N KO, V. S . G O R NA KOV, L . M. D E D U K H,Y U.P. K A BA N OV, A. F. K H A P I KOV, A. J . S H A P I RO,R . D. S H U L L, A. C H A I K E N and R. P. M I C H E L , Phys. Rev.B57 (1998) R8111.

26. A . P. M A L O Z E M O F F , Phys. Rev. B37 (1988) 7673; N . C .KO O N , Phys. Rev. Lett. 78 (1997) 4865; T. C . S C H U LT H E S S

and W. H. B U T L E R , ibid. 81 (1998) 4516; R . L . S TA M P S , J.Phys. D: Appl. Phys. 33 (2000) R247.

27. P. B L O M Q V I S T, K A N NA N M. K R I S H NA N, H. O H L DAG,A. S C H O L L and J . S T O H R , Phys. Rev. Lett. 94 (2005) 107203.

28. L . N E E L , J. Phys. Radium 15 (1954) 225.29. N . -H. C H O, K A N NA N M. K R I S H NA N, C. L U-

C A S and R. F. C . FA R ROW , J. Appl. Phys. 72 (1992)5799.

30. M. Z I E S E and M. J . T H O R N TO N (eds.), “Spin Electronics”(Springer-Verlag, Berlin, Heidelberg, New York, 2001).

31. M. N. BA I B I C H, J . M. B ROTO, A. F E RT, F. N .VA N DAU, F. P E T RO F F, P. E I T E N N E, G. C R E U Z E T, A.F R I E D E R I C H and J . C H A Z E L A S , Phys. Rev. Lett. 61 (1988)2472.

32. V. S . S P E R I O S U, B . D I E N Y, P. H U M B E RT, B . A.G U R N E Y and H. L E FA K I S , Phys. Rev. B44 (1991) 5358.

33. J . S . M O O D E R A, L . R . K I N D E R, T. M. W O N G and R.M E S E RV E Y , Phys. Rev. Lett. 74 (1995) 3273.

34. P. NA J I , M. D U R L A M, S . T E H R A N I , J . C A L D E R andM. D E H E R R E R A , 2001 IEEE International Solid State CircuitsConference, (2001) p. 437.

35. G . S C H M I D T and L . W. M O L E N K A M P , Physica E 10 (2001)484.

36. Y. O H N O, D. K. YO U N G, B. B E S C H OT E N, F. M AT-S U K U R A, H. O H N O and D. D. AW S C H A L O M , Nature(London) 402 (1999) 790.

37. S . A . W O L F, D. D. AW S C H A L O M, R. A. B U H R M A N,J . M. DAU G H TO N, S . VO N M O L NA R, M. L . RO U K E S,A. Y. T C H E L K A N OVA and D. M. T R E G E R , Science 294(2001) 1488.

38. K . A . G R I F F I N , A. B . PA K H O M OV, C. M. WA N G, S .M. H E A L D and K A N NA N M. K R I S H NA N , Phys. Rev. Lett. 94(2005) 157204.

39. L . R E I M E R , “Transmission Electron Microscopy” (Springer-Verlag, 1997).

40. M. VA R E L A, W. G RO G G E R, D. A R I A S , Z . S E F R I O U I ,C . L E O N, C . BA L L E S T E RO S, K A N NA N M. K R I S H NA N

and J . S A N TA M A R I A , Phys. Rev. Lett. 86 (2001) 5156.41. J . S A N TA M A R I A, M. E . G O M E Z, J . L . V I C E N T,

K A N NA N M. K R I S H NA N and I . S C H U L L E R , ibid. 89 (2002)190601.

42. N . C H E N G, K A N NA N M. K R I S H NA N, E R. G I RT, R . F.C . FA R ROW, R. M A R K S, A. YO U N G and H. A. C H E N G ,J. Appl. Phys. 87 (2000) 6647.

43. G . D E N B E AU X, P. F I S C H E R, G. K U S I N S K I , M. L E

G RO S, A. P E A R S O N and D. AT T W O O D , IEEE Trans. Magn.37 (2001) 2764.

44. A . S C H O L L, H. O H L DAG, F. N O LT I N G, J . S T O H R

and H. A. PA D M O R E , Rev. Sci. Instrum. 73 (2002)1362.

813

40TH ANNIVERSARY

45. P. TA RTA J , Encycl. Nanosc. Nanotech. 6 (2004) 823.46. V. F. P U N T E S , K A N NA N M. K R I S H NA N and P. A .

A L I V I S ATO S , Science 291 (2001) 2115.47. Y. BAO, A. B . PA K H O M OV and K. M. K R I S H NA N , J. Appl.

Phys. 97 (2000) 10J317.48. V. K . L A M E R and R. H. D I N E G A R , J. Am. Chem. Soc. 72

(1950) 4847.49. Y. BAO, M. B E E R M A N, A. B . PA K H O M OV and K. M.

K R I S H NA N , J. Phys. Chem. B. 109 (2005) 7220.50. Y. BAO, M. B E E R M A N and K. M. K R I S H NA N , J. Mag.

Mag. Mat. 266 (2003) L245.51. T. C . L U B E N S K Y , Solid State Commun. 102 (1997) 187.52. W. Z H O U, A. K U M B H A R, J . W I E M A N N, J . FA N G, E .

E . C A R P E N TA R and C. J . O’C O N N E R , J. Solid State Chem.159 (2001) 26.

53. S- J . C H O, S . M. K AU Z L A R I C H, J . O L A M I T, K. L I U ,F. G R A N D J E A N, L . R E B B O U L H and G. J . L O N G , J. Appl.Phys. 95 (2004) 6804.

54. J . L . LYO N, D. A. F L E M I N G, M. B . S TO N E, P.S C H I F F E R and M. E . W I L L I A M S , Nanoletters 4 (2004) 719.

55. A . K . G U P TA and M. G U P TA , Biomaterials 26 (2005) 3995.56. Y. BAO and K. M. K R I S H NA N , J. Magn. Magn. Mater. 293

(2005) 15.57. S . M O R N E T, S . VA S S E U R, F. G R A S S E T and E . D U G U E T ,

J. Mater. Chem. 14 (2004) 2161.58. J . C . R I F E , M. M. M I L L E R, P. E . S H E E H A N, C. R .

TA M A NA H A, M. TO N D R A and L. J . W H I T M A N , Sens.Actuators A, Phys. A107 (2003) 209.

59. L . E J S I N G, M. F. H A N S E N, A. K. M E N O N, H. A.F E R R E I R A, D. L . G R A H A M and P. P. F R E I TA S , Appl. Phys.Lett. 84 (2004) 4729.

60. A . B . PA K H O M OV, Y. BAO and K. M. K R I S H NA N , J. Appl.Phys. 97 (2005) 10Q305.

61. J . C O N N O L LY and T. G. S T. P I E R R E , J. Magn. Magn. Mater.225 (2001) 156.

62. A . J O R DA N , in Scientific and Clinical Applications of MagneticCarriers, edited by U. Hafeli (Plenum Press, New York, 1997).

63. R . H E R G T and W. A N D R A , IEEE Trans. Magn. (USA) 34 (1998)3745.

64. M. G O N Z A L E S and K. M. K R I S H NA N , J. Magn. Magn. Mater.293 (2005) 265.

65. I . A . B R E Z OV I C H , Med. Phys. Monograph 16 (1988) 82.66. R . E . RO S E N S W E I G , J. Magn. Magn. Mater. 252 (2002) 370.67. P. C . FA N N I N, B . K. P. S C A I F E and S . W. C H A R L E S ,

ibid. 72 (1988) 95.68. J . A . C . B L A N D and B. H E I N R I C H (eds), “Ultrathin Magnetic

Structures I & II” (Springer-Verlag, 1994).69. L . PA L, K. K R E N, G. K A DA R, P. S Z A B O and T.

TA R N O C Z I , J. Appl. Phys. 39 (1968) 538.70. N . C H E N G, J . P. A H N and K A N NA N M. K R I S H NA N , ibid.

89 (2001) 6597.71. P. B L O M Q V I S T, K A N NA N M. K R I S H NA N and D. E .

M C C R E A DY , ibid. 95 (2004) 8019.72. L . G . PA R R AT , Phys. Rev. 95 (1954) 359.73. P. B L O M Q V I S T and K A N NA N M. K R I S H NA N , J. Appl. Phys.

95 (2004) 8487.74. E . C . S TO N E R and E . P. W O H L FA RT H , Proc. Phys. Soc.

A240 (1948) 599.75. P. B L O M Q V I S T, K A N NA N M. K R I S H NA N, S . S R I NAT H

and S . G . E . T E V E LT H U I S , J. Appl. Phys. 96 (2004) 6523.76. A . S C H O L L, H. O H L DAG, F. N O LT I N G, J . S T O H R, H.

A. PA D M O R E , Rev. Sci. Instrum. 73 (2002) 1362.77. P. B L O M Q V I S T, K A N NA N M. K R I S H NA N and H.

O H L DAG , Phys. Rev. Lett. 94 (2005) 107203.78. C . L . C A N E DY, X. W. L I and G. X I AO , J. Appl, Phys. 81

(1997) 5367.

79. A . B E R G E R and H. H O P S T E R , Phys. Rev. Lett. 76 (1996) 519.80. C . KO OY and U. E N Z , Philips Research Reports 15 (1950) 7.81. J . P. Q I A N and G. C . WA N G , J. Vac. Sci. Technol. A8 (1990)

4117.82. Y. S . C H U N and K A N NA N M. K R I S H NA N , J. Appl. Phys. 95

(2004) 6858.83. M. J . D O NA H U E and D. G. P O RT E R , Interagency Report

NISTIR 6376, OOMMF User’s Guide, Version 1.0. (1999).84. L . D . L A N DAU and E. L I F S H I T Z , Physik. Z. Sowjet Union 8

(1935) 153.85. J . S L O N C Z E W S K I , J. Magn. Magn. Mat. 150 (1995) 13.86. A . RO S E N C WA I G, W. J . TA B O R and R. D. P I E R C E , Phys.

Rev. Lett. 26 (1971) 779.87. B . T. J O N K E R, S . C . E RW I N, A. P E T RO U and A. G.

P E T U K H OV , MRS Bulletin 28 (2003) 740.88. Y. M AT S U M OTO, M. M U R A K A M I , T. S H O N O, T.

H A S E G AWA, T. F U K U M U R A, M. K AWA S A K I , P.A H M E T, T. C H I K YOW, S . -Y. KO S H I H A R A and H.KO I N U M A , Science 291 (2001) 854.

89. S . A . C H A M B E R S, S . T H E V U T H A S A N, R . F. C .FA R ROW, R. F. M A R K S, J . U . T H I E L E, L . F O L K S,M. G. S A M A N T, A. J . K E L L O C K, N. RU Z Y C K I , D . L .E D E R E R and U. D I E B O L D , Appl. Phys. Lett. 79 (2001) 3467.

90. K . U E DA, H. TA BATA and T. K AWA I , ibid. 79 (2001) 988.91. S . K U WA BA R A, T. KO N D O, T. C H I K YOW, P. A H M E T

and H. M U N E K ATA , Jpn. J. Appl. Phys. Part 2 40 (2001)L724.

92. M. K. R E E D, N. A. E L-M A S RY, H. H. S TA D E L M A I E R,M. K. R I T U M S, M. J . R E E D, C . A. PA R K E R, J . C .RO B E RT S and S . M. B E DA I R , Appl. Phys. Lett. 79 (2001) 3473.

93. N . T H E O D O RO P O U L O U, A. F. H E BA R D, M. E . OV E R-B E R G, C . R . A B E R NAT H Y, S . J . P E A RTO N, S . N . G.C H U and R. G. W I L S O N , Phys. Rev. Lett. 89 (2002) 107203.

94. S . G . YA N G, A. B . PA K H O M OV, S . T. H U N G and C. Y.W O N G , Appl. Phys. Lett. 81 (2002) 2418.

95. Y. W U, H. X. L I U , L . G U, R . K. S I N G H, L . B U D D,M. VA N S C H I L F G A A R D E, M. R . M C C A RT N E Y, D. J .S M I T H and N. N E W M A N , ibid. 82, (2003) 3047.

96. S . B . O G A L E, R . J . C H O U D H A RY, J . P. B U BA N, S .E . L O F L A N D, S . R . S H I N D E, S . N . K A L E, V. N.K U L K A R N I , J . H I G G I N S , C . L A N C I , J . R . S I M P S O N,N. D. B ROW N I N G, S . DA S S A R M A, H. D. D R E W, R.L . G R E E N E and T. V E N K AT E S A N , Phys. Rev. Lett. 91 (2003)077205.

97. J . M. D. C O E Y, A. P. D O U VA L I S , C . B . F I T Z G E R A L D

and M. V E N K AT E S A N , Appl. Phys. Lett. 84 (2004) 1332.98. S . R . S H I N D E et al., Phys. Rev. B67 (2003) 115211; J . -Y. K I M,

J . H . PA R K, B . G. PA R K, H. J . N O H, S . J . O H, J .S . YA N G, S . D . B U, T. W. N O H, H. J . L I N , H . H.H S I E H and C. T. C H E N , Phys. Rev. Lett. 90 (2003) 017401; S .A . C H A M B E R S E T A L. , Thin Sol. Films 418 (2002) 197; S .A . C H A M B E R S, S . M. H E A L D and T. D RO U BAY , Phys.Rev. B67 (R) (2003) 100401; S . A . C H A M B E R S and R. F. C .FA R ROW , MRS Bulletin 28 (2003) 729; G . C . H A N et al., J.Magn. Magn. Mater. 268 (2004) 159.

99. M. VA R E L A, K. A. G R I F F I N , S . P E N N Y C O O K andK A N NA N M. K R I S H NA N (unpublished).

100. RBS measurements courtesy of V. Shutthanandan, Pacific NorthwestNational Laboratory, Richland, WA.

101. M. S . PA R K, S . K . K W O N and B. I . M I N , Phys. Rev. B65(2002) 161201R.

102. K . A . G R I F F I N , A. B . PA K H O M OV, C. M. WA N G, S .M. H E A L D and K A N NA N M. K R I S H NA N , J. Appl. Phys. 97(2005) 10D320 (Note: Undoped TiO2 prepared and annealed underthe same conditions is conducting.)

103. A . B . PA K H O M OV, B. K. RO B E RT S and K. M. K R I S H-

814

40TH ANNIVERSARY

NA N , Appl. Phys. Lett. 83 (2003) 4357.104. A . B . PA K H O M OV, B. K. RO B E RT S , V.

S H U T T H A NA N DA N, D. M C C R E A DY, S . T.T H E V U T H A S A N, A. T UA N, S . A . C H A M B E R S andK. M. K R I S H NA N , J. Appl. Phys. 95 (2004) 7393.

105. P. S H E N G , Philososphical Magazine B65 (1992) 357.106. B . K . RO B E RT S , A. B . PA K H O M OV, V. S . S H U T-

T H A NA N DA N and K A N NA N M. K R I S H NA N , J. Appl. Phys.97 (2005) 10D310.

107. See http://depts.washington.edu/kkgroup for further details.

815