parkin annurev

8/12/2019 Parkin Annurev

1/34

etnnu. Rev. Mater. Sci. 1995.25. 357-88Copyright 1995 by Annual Reviews 1no. All rights reserved

GIANT MAGNETORESISTANCE INMAGNETIC NANOSTRUCTURESS. S. P. Park&IBM Research Division, Almaden Research Center, 650 Harry Road,San Jose, California 95120KEYWORDS:multilayers, granular metals, interlayer exchange oupling,magnetic ecording, sputtered films

ABSTRACTThis chapter contains a brief review of the giant magnetoresistance (GMR)effect exhibited by magnetic multilayers, granular alloys, and relatedmaterials. Subjects covered include a description of the phenomenon, ndthe related oscillatory interlayer exchange coupling in magnetic multi-layers; a simple model of giant magnetoresistance; the inverse GMRffectin spin-engineered magneticmultilayers; structures that display large changesin resistance in small magnetic fields, possibly for use in magnetic fieldsensors; and the dependence of GMR n various aspects of the magneticstructures.

INTRODUCTIONThere has been much nterest in recent years in artificially engineerednano-structured materials with novel physical properties. One area ofparticular interest is that of metal multilayers. These materials have beenstudied for the past 30 years or more, but it is only relatively recentlythat detailed expertise has been developed to prepare and sufficientlycharacterize the structure of such materials (1-3). Metal multilayers havebeen studied for their novel superconductingproperties, for the possibilityof creating metals much stronger and tougher than the individual com-ponents of the multilayer, for creating super-mirrors for X-ray or neutronscattering monochromators, nd for a host of other reasons. Most recently,interest in metal multilayers has centered on multilayers comprised of

0084-6600/95/0801-0357505.00 357

www.annualreviews.org/aronlineAnnual Reviews

Annu.Rev

.Mater.Sci.1995.25:357388.D

ownloadedfromarjournals.annu

alreviews.org

byMax-Planck-Gesellschafton05/06/05.

Forpersonaluseonly.
http://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronlinehttp://www.annualreviews.org/aronline


2/34

358 PAR~Ythin magnetic layers separated by thin nonferromagnetic layers. Thesemultilayers display unusual magnetic and transport properties that are thesubject of this brief review. Note that several excellent comprehensivereviews of the transport and magnetic properties of magnetic multilayersand related materials are available (4-10).Usually the change in resistance of thin metal films with application ofa magnetic field is small (11). However, n 1988 it was found that singlecrystalline (100)-oriented Fe/Cr/Fe sandwiches (12) and (100)-orientedFe/Cr multilayers (13) displayed much larger magnetoresistance valuesthan could be accounted for by the magnetoresistance of the individualFe layers themselves. The resistance of the multilayers was found todecrease by almost a factor of two when a field of ~ 20 kOe was appliedat low temperatures (4.2 K) (13). The term giant magnetoresistance (GMwas coined to describe this effect. The thickness of the Cr layers in thesestructures was chosen to correspond to the thickness that was previouslyfound to give rise to an unusual antiferromagnetic interlayer exchangecoupling of neighboring Fe layers (14, 15). These films were preparedmolecular beam epitaxy (MBE)(see, for example, 16-18), whichsophisticated and expensive ultra-high vacuumdeposition technique. Soonit was shown hat similar GMResults could be obtained in polycrystallineFe/Cr sandwiches and multilayers prepared using inexpensive and simplersputter-deposition methods (19, 20). Sputter-deposition allowed for thestudy of a wide variety of metal multilayers, and this led to the surprisingdiscovery that GMR as commono many metal multilayers (19, 21).date, the largest GMRffects at room temperature are found in Co/Cumultilayers comprised of thin Co and Cu layers (21-23).Interest in transition metal multilayers has also been greatly increasedin recent years with the discovery of oscillations in the interlayer magneticexchange coupling in multilayers comprised of thin ferromagnetic layersof Fe, Co, Ni, and their various alloys separated by thin layers of virtuallyall the nonferromagnetic transition and noble metals (19, 22, 24, 25).This discovery led to the prospect of artificially engineering multilayeredstructures with complex magnetic structures, useful both for under-standing the physics of these materials as well as for technological appli-cations. Indeed, magnetic thin film structures already have found impor-tant applications for storing information in magnetic and magneto-opticaldisk storage devices (26, 27), for magnetic field sensors in magnetic re-cording magnetoresistive read heads (28), and for nonvolatile randomaccess memory (29). Applications involving hybrid magnetic/semi-conductor structures have been proposed (30), and recently, spin switches,devices based on transport of spin-polarized electrons across magnetic/nonmagnetic interfaces, have been devised and prototypes built (31, 32).


Annu.Rev.Mater.Sci.1995.25:357388.Downloadedfromarjournals.annualreviews.org

byMax-Planck-Gesellschafto

n05/06/05.

Forpersonaluseonly.


3/34

GIANTMAGNETORESISTANCE59It is likely that devices based on the giant magnetoresistance henomenonwill be used withina fewyears.

GIANT MAGNETORESISTANCEMagnetic Multilayers: Polycrystalline Co/Cu and Fe/CrFigure 1 showsplots of the change n room emperature esistance ARvsin-plane magnetic ield for two polycrystalline multilayers of Fe/Cr and

Co/Gu513 .I

,,,,,, ..... Fe/Cr ................................................,0"o

,,, H(kOe)-40 -20 0 20 40

Figure 1 Roomemperature resistance vs in-plane magnetic field curves for polycrystallineFe/Cr and Co/Cu multilayers deposited by magnetron sputtering (for detailed depositionconditions, see 19, 21, but note that the Co/Cumultilayer was deposited at 2.1 mTorr). Themeasurementgeometry is shown schematically in the top left corner. The magnetic state ofthe antiferromagnetically coupled multilayers is shown chematically in the lower portion ofthe figure for large negative, zero, and large positive magnetic ields.


Annu.Rev.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly

.


4/34

360 PARKINCo/Cu, espectivelyI. Note hat the resistance is measuredwith the currentin the plane of the layers, and he magnetic ield is applied orthogonal othe current also in the plane of the layers. At high magnetic ields, theresistance saturates at somevalue R, and the sampleresistance is nor-malized o this value. The igure showshat the resistance of the multilayersis significantly higher in small fields as compared ith high fields. Thevariation in resistance is related to a changen the relative orientation ofneighboringferromagnetic Fe or Co layers with applied magnetic field.Theresistance is higher when djacent magnetic ayers are aligned anti-parallel to one another, as comparedwith parallel alignment, shown che-matically in cartoons of the magneticmultilayer included in Figure 1. Inthe figure, values of room-temperatureaturation magnetoresistance XR/Rof more han 25 and 70% re observed or Fe/Cr and Co/Cu, espectively,with corresponding saturation fields of --,25 and 10 kOe. The cor-respondingGMRalues at 4.2 K are significantly higher at 110 and 130%,respectively.The GMRf the samples shown n Figure 1 are some of the highestrecorded GMR alues in magnetic multilayers. Higher values still havebeen reported in single crystalline (100) Fe/Cr multilayers. GMRaluesexceeding 150% t 4.2 Khave been found in magnetron-sputtered amples(33), and values of more han 220% t 1.5 K have been reported recentlyin MBE-deposited ultilayers (34).As shown chematically n Figure 1, the magneticmomentsf successivemagnetic layers in the Co/Cu nd Fe/Cr multilayers are arranged anti-parallel to one another n small fields. This is a consequence f an anti-ferromagnetic interlayer exchangecoupling J,~F propagated hrough theinterveningCr or Cu ayers. As he Cr or Cu ayer thickness is varied, theexchange oupling of the magnetic ayers is found to vary in sign, oscil-lating between antiferromagnetic (AF) and ferromagnetic (F) coupling(19, 22). This is manifested, or example, s an oscillation in the magnitudeof the GMRffect with increasing separation of the magnetic ayers and

~ Thestructures of the Fe/Cr and Co/Cumultilayers shownn Figure 1 are as follows:Si/40/~Cr/[8/~Fe/7.5/~Cr]39/8/~Fe/15/~Cr, nd Si/50/~Fe/[8/~Co/7.5~Cu]59/8/~Co/20~Fe.Bothmultilayers weredepositedon Si(100), whichwascoveredwith a thin (- 10-15~, thick)native oxide layer. TheFe/Cr multilayer was deposited at 125C,whichwasfound o giverise to maximalGMRalues (20). It waspostulated that this wasa result of reducing hebulk resistivity of the Fe and Cr layers while imiting intermixingof the Fe andCr layers.TheCo/Cumultilayer wasdeposited at room emperaturebecause deposition at any highertemperatureappears to decrease GMR,resumably ecause of dissolution or intermixingofthe Coand Cu ayers (21-23). AnFe buffer layer is used. It wasdiscovered hat Fe bufferlayers give the largest GMRalues in Co/Cumultilayers grown n silicon for thin Cu ayers(21,22).




n05/06/05.

Forpersonaluseonly.


5/34

GIAN T MAGNETORESISTANCE 361was first observed in sputtered Fe/Cr multilayers 19). Figure 2 givesresults for a series of Co/Cu multilayers in which the magnitude of thesaturation magnetoresistance is found to oscillate with increasing Cuspacer layer thickness with an oscillation period of ~9 8,. Large GMRvalues are found for Cu layer thicknesses for which the Co layers arecoupled antiferromagnetically see Figure 2). For strong ferromagneticcoupling of the Co layers, the relative magnetic alignment of the Co layersis unaffected by magnetic field, and consequently there is no GMR effect.Note that for polycrystalline multilayers, as shown in Figure 2, a significantbackground magnetoresistance MR) effect is superimposed on the GMRoscillations. This effect is most likely a consequenceof the crystallographicstructure of the multilayer, which is comprised of small crystalline grains,typically 100 to 200 8, in size (35). Although the grains are preferentiallyoriented along ( l l l ) , the texture is poor, and there is a considerableangular dispersion of the (1 1 1 axis from grain to grain. In addition, theremay also be 100)-oriented grains (36). Because the period of the oscillatorycoupling in Co/Cu depends on crystallographic orientation (37, 38), it ispossible that CMR oscillations from differently oriented grains in thepolycrystalline Co/Cu structures could be superimposed, perhaps account-ing for the G M R background shown in Figure 2. For very thick Cu layers,where the interlayer coupling is very weak, GMR is still observed if themagnetic layers break u into magnetic domains, as shown schematically

Figure 2 Room temperature saturation magnetoresistance vs Cu spacer layer thickness fora series of polycrystalline sputter-deposited CojCu multilayers (from 22).Th e m agnetic stateof the multilayer is shown schematically for various Cu layer thicknesses (only two magneticlayers are shown).


Ann

u.Rev.Mater.Sci.1995.25:357388.Downloadedfromarjournals.annualreviews.org

byMax-Planck-Gesellsch

afton05/06/05.

Forpersonaluseonly.


6/34

362 PARKINin Figure 2. GMRs the result of electrons propagating rom one magneticlayer to another magnetic ayer for which he magneticmoments re notaligned with one another. Thus here is a certain probability that electronsemanating rom a magnetic domain n one magnetic ayer will propagateand undergoscattering in a magneticdomain n a neighboring ayer (orindeedeven n the same ayer), whichwould ead to an increased esistance.The resistance Js highest in this case when he net magneticmoment feach ayer is close to zero at the coercive ield He. This leads to twin peaksin resistance t .+_Granular Alloys--An Example. [l l l]-Oriented CoCuThe preceding discussion of GMRn uncoupled Co/Cumultilayers indi-cates that magnetically oupledmultilayers are not a prerequisite for theobservation of GMR. he layered geometry is also unnecessary, andgiant magnetoresistancehas been observed in a variety of inhomogeneousmagnetic systems. The only requirement is that the system contains adistributed magneticcomponentn which he relative orientation of themagnetic moments f small magnetic entities will vary with magneticfield or any other variable (such as, for example, emperature r strain).Granular alloys containing magneticparticles in a metallic host bear astrong resemblance o magneticmultilayers, and GMRas been observedin a variety of magneticgranular systems predominantly omprisedof Fe,Co, Ni, and their various alloys in Cu, Ag, and Aumatrices (39-42) (fora reviewsee C L Chien, this volume).Figure 3 shows ypical GMResultsfor a single crystalline (111)-orientedC028Cu72ranular alloy preparedMBE41). Resistance vs field curves are shown or the magneticfieldapplied both in-plane and orthogonal to the sensing current and per-pendicular to the plane of the film. There is a significant magneticanisotropy, perhaps resulting from the shapeof the Co particles, or fromthe intrinsic magneticanisotropy of the Co itself. Note also that theresistance of the sample s dependent n the magnetic ield history. As aresult of the considerablemagnetic nisotropy, he coercivefields are verydifferent for field applied in the plane of and perpendicularo the plane ofthe samples.Just as for the uncoupledCo/Cumultilayers described above,the largest resistance in granular alloys is found when he magnetizationis close to zero at +_H~.This corresponds o the state of maximal nti-alignment of the magnetic moments f neighboring magnetic particles.Thus, because he coercive field dependson field orientation, the sampleresistance in zero field can be varied by up to a factor of two, as shownnFigure 3. In this case, the samplewasfirst magnetized erpendicular othe film. After reducing he field to zero, the samplewasrotated so thatthe field wasaligned in the plane of the sample. Magnetizinghe sample




n05/06/05.

Forpersonaluseonly.


7/34

GIANT MAGNETORESISTANCE 363

ield Oe)

Figure 3 Resistance vs field curves for a M BE-deposited crystalline (1 11)-oriented C O U ~ ~alloy. The film was deposited at 2 C and annealed for 1 h at 3 C (for depositionconditions, see 41). Data are shown at 4.2 K for field applied in-plane (and orthogonal tothe sense current) and for field applied along the field normal. The arrangement of themagnetic Co particles is shown schematically in the bottom portion of the figure for largenegative, zero, an d large positive magne tic fields.

and reducing the field to zero results in an increase of the zero fieldresistance by 5 1YOOSCILLATORY INTERLAYER C OU PLINGThe Cu or Cr in the layers adjacent to the Co or Fe magnetic layers shownin Figure 1 becomes spin-polarized, and these atoms develop magneticmoments. For Cu, the d-spin-induced moment, which is parallel to the Comoment, is very small and is about 100 times smaller than the moment onCo itself, as deduced from X-ray magnetic circular dichroism studies(XMCD) 43). For Cr, the induced d-moment, which is antiparallel to theFe moment 44), is significantly larger and is comparable to that of theFe moment 45).The induced spin-polarization of Cr has been directlyobserved by measuring the spin-polarization of secondary electrons excited


Annu.Re

v.Mater.Sci.1995.25:357388.Downloadedfromarjournals.annualreviews.org


n05/06/05.

Forpersonaluseonly.


8/34

364 PARKINfrom a (100)-oriented single crystalline Cr wedge eposited on a singlecrystal Fe whiskerby the electron beam n a scanningelectron microscope(SEMPA)10, 44). Thedirection of the induced Cr momentt the surfaceof the Cr layer is observed o oscillate with increasing Cr thickness in amanner onsistent with the variation of the interlayer magnetic ouplingbetweenFe layers observed n related structures in whicha thin Fe layeris deposited on top of the Cr wedge 46). It is the induced spin densitywave n the spacer layer material that mediates he magneticcoupling ofthe magnetic ayers in magneticmultilayers and sandwiches.The nterlayer excl" nge couplingof thin layers of Fe, Co, Ni, and theiralloys via nonmagneticransition and noble metal spacer layers nearlyalwaysexhibits an oscillatory variation with spacer layer thickness (25).Theoscillation period varies frommetal to metal and lies between ~ 8 ,~and 12 ~ in mostcases, with the exception f Cr, for which t is significantlylonger (-~ 18 ,~). In some ases, more han one oscillation period has beenobserved. For Cr, a secondshort oscillation period just -~ 2 monolayerslong has beenobserved 46, 47). The nterlayer couplinghas be,en exploredexperimentallyn many ystemsusing a widevariety of techniques, includ-ing SEMPAnd XMCD,s well as more conventional techniques includ-ing various magnetometries, rillouin light scattering (see, for example,48-50) and ferromagnetic esonance 51). A detaileddiscussion of inter-layer couplingcan be found elsewhere 7, 10, 38, 52-54).A SIMPLE MODEL OF GIANTMAGNETORESISTANCEThe detailed origin of GMRas provoke.d considerable interest. Manytheoretical modelshave been developed, but most of themare based on amodel f the electrical conduction n ferromagneticmetals fromMott(55).Mott hypothesized hat the electrical current in ferromagneticmetals iscarried independently in two conduction channels that dorrespond pre-dominantly o the spin-up and spin-down -p electrons. These electronsare in broad energy bands with low effective masses. This assumption sbelieved to be good at temperatures significantly below the magneticordering emperature f the magneticmaterial so that there is little spin-mixing between the two conduction chhnnels. Mott theorized that theconductivity an be significantly different in the twospin channelsbecausethe conduction-electron scattering rates in these two channels will berelated to the corresponding pin-up or spin-down ensity of empty tatesat the Fermi evel. Thesestates will be largely of d character, and as aresult of the exchange plit d bands, the ratio of spin-up to spin-downdensity of empty tates at the Fermi evel can be significantly different




n05/06/05.

Forpersonaluseonly.


9/34

GIANT MAGNETORESISTANCE 365in the ferromagnetically ordered states of Fe, Ni, Co, and their alloys.Consequently, his leads to the possibility of substantially different mean-free paths 2 and conductivities a-+ in the two channels. In Co, for example,the density of states at the Fermi level is ten times higher for down-spin(minority) electrons as comparedwith up-spin (majority) electrons (56).(For in-depth reviews see, for example, 55-58.)

Detailed theories of the giant magnetoresistance effect in magneticmultilayers have been developed (52, 59-66). However, it is beyond thescope of this short review to discuss these models n any detail, but see thedetailed discussion in reference (8). The simplest model for a magneticmultilayer that results in GMRs an equivalent resistor network model(53, 64), shown schematically in Figure 4. In this model, each of theferromagnetic and nonmagnetic spacer layers consists of two resistorscorresponding to the two conductivity channels associated with the up-and down-spin lectrons. In the ferromagnetic ayers, the resistivity is spin-dependent, p~, whereas in the spacer layers, the resistivity in the twochannels s identical, Ps. The resistance of the multilayer is then equivalentto that of a total of eight resistors, with four resistors in each channel. Thenet resistivities of the two channels can be treated as resistors in parallel.Appropriately summing he resistors within a given channel is more com-plicated, but there are two simple cases (53). For short mean-free pathscomparedwith the thickness of the layers, the resistors are independentand should themselves be added in parallel. Under hese circumstances, itis clear that the resistance in the ferromagnetic and antiferromagneticconfigurations is the same, and consequently there is no magnetoresistancedefined as AR/R = (RAF--RF)/RF, where RAF and RF are the resistancescorresponding to the AF and F configurations. Another straightforwardcase is when the mean-free paths are long comparedwith the layer thick-nesses in the multilayer. Then he resistivity is an averageof the resistivityof the various layers in the multilayer, in proportion to the thicknesses ofthe corresponding layers. Note that for the F configuration, only tworesistivities must be averaged, but in the AFconfiguration there are fourresistivities to be considered. Taking these averages, and subsequentlyadding the resistivities of the two spin channels in parallel, leads toAR/R= [(~-fl)2]/[4(~-b N/M)(fl N/M)], where N and M are the thick-nesses of the spacer and ferromagnetic layers, and ~ = P~/Ps andfl = PY/Ps. The magnetoresistance in this model depends on twoparameters, ~z/fl and Mfl/N. This model shows, not surprisingly, that themagnitude of AR/R is strongly dependent on the scattering asymmetrybetween the spin conduction channels in the ferromagnetic layers. Ofcourse it is irrelevant in whichspin channel the scattering is stronger, andthe magnitude of the MR epends on howmuch he ratio ~/fl differs from




alreviews.org


Forpersonaluseonly

.


10/34

366 PARKIN

Parallel nti parallel

UPspin

downspin

UPspin

downspinFigure 4 A simple resistor network model of GMR in which the height of the columns isproportional to the resistivity of the magnetic arrows)and spacer layer white boxes metalsfor the two independent up- and down-spin channels. The magnetic orientation of themagnetic layers is shown by the direction of the arrow inscribed in the relevant box. Twodifferent multilayers are considered. In the top portion of the figure, a simple multilayercontaining one type of magnetic layer is described. In the bottom po rtion, a resistor networkmodel is shown for a magnetic multilayer containing two different types of magnetic layersseparated by the same spacer layer.


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


n05/06/05.

Forpersonaluseonly.


11/34

GIANT MAGNETORES1STANCE 3671. This highlysimplified model lso predicts that for a constant atio e///,the MR ecreases monotonicallywith increasing spacer layer thickness,falling off as 1IN2 for large N. The MRs actually found to decreaseexponentiallywith N for large N (67). Thereason for this discrepancythat the resistor networkmodel s no longer applicable for N arge com-pared to the mean-freepath in the spacer layer. Sucha simple resistornetworkmodel an easily give values of MRxceeding100%or ~//3 ratiosof -~ 8 to 10 (13, 53), and these ratios are considered easonable 56, 57).Nevertheless, we note that, as shown y the resistor networkmodel, hemagnitude f the MRs expectedo be related to the ratio of the scatteringrates within the two conduction hannels no matter where he spin-depen-dent scattering takes place. The scattering asymmetries have beenindirectly determined rom measurements f the resistivity of magneticternary alloys (56-58). However, o correlation between he magnitudethe scattering asymmetries romstudies of bulk magneticalloys and themagnitude f the MRn magneticmultilayers has yet been found.Thesimple resistor networkmodeldiscussed above predicts that theresistance of a magneticmultilayerwill be higherfor antiparallel alignmentof the magnetic ayers as compared ith parallel alignment. This is sche-matically demonstratedn the upperportion of Figure 4, where t is clearthat the lower resistance for parallel alignmentof the magnetic ayersresults from a short circuit of the current throughone of the spin con-duction channels shown n the figure as the down-spin hannel). A simpleextension of this model can be made or the case of a ferromagneticmultilayer containing two different ferromagnetic ayers with differentratios (~//~), shown n the cartoon n the lowerpart of Figure 4. In thiscase, the resistance of the multilayer will be lowerfor the antiparallelalignmentof neighboringmagneticayers, thereby eading to a positive oran inverse giant magnetoresistance ffect. This effect has not yet beenunambiguouslyemonstrated,but the observation of a large inverse GMReffect wouldbe important to clarify theoretical modelsof GMR.INVERSE GIANT MAGNETORESISTANCE IN SPIN-ENGINEERED MULTILAYERSThe interlayer exchange coupling strength in antiferromagneticallycoupled magnetic multilayers can be readily measuredfrom the fieldrequired to rotate the magneticmoments f neighboringmagnetic ayersparallel to one another(68-71). Indeed, the strength of the AF nterlayercouplingJAY s related to the magnetizationMsand thickness tv of themagnetic ayers and the field Hs needed o rotate the magneticmomentsof these layers parallel to one another, by the relation, JAy~-- HsMstF/4




alreviews.org


Forpersonaluseonly

.


12/34

368 PARKIN(71, 72). It is moredifficult to determineferromagnetic coupling strengths,although these can be calculated from the stiffness of spin wavemodes nferromagnetic resonance (73) or Brillouin light-scattering studies (50).simple technique to measure ferromagnetic coupling strengths in multi-layers was developed by spin-engineering appropriate structures such thatthe coupling strength could be inferred directly from the magnetization vsfield loop (24). The method s shown n the upper portion of Figure 5. Thebasic idea is to ensure that the two magnetic layers of interest (the topand middle Co layers in the example shown in Figure 5), which areferromagnetically coupled, have their magnetic momentsarranged anti-parallel to one another for some ield range. In Figure 5, the two uppermostCo layers are weakly ferromagnetically coupled through a spacer layer ofCu. A third Co layer (the bottom layer in the figure) is designed tostrongly antiferromagnetically coupled to the middle Co layer througha thin layer of Ru. It turns out that Co layers are coupled stronglyantiferromagnetically for Ru layers a monolayer or so thick (19). Thusstructure has been spin-engineered in which at low fields the top twoferromagnetically coupled Co layers are aligned antiparallel to the appliedfield. This occurs because the magnetic moment f the bottom Co layer isdesigned to be slightly larger than the sum of the magnetic momentsofthe top two Co layers. Because the ferromagnetic coupling 6f the toptwo layers is muchweaker in strength than the strong antiferromagneticcoupling between the bottom two Co layers, when a magnetic field isapplied, the magnetic moment f the topmost Co layer rotates in the fieldand becomesaligned with the field for intermediate field strengths. Thus,as shown n Figure 5, the two ferromagnetically coupled Co layers becomealigned antiparallel to one another. The field required to rotate the toplayer in this way will be proportional to the ferromagnetic interlayercoupling strength (provided the middle Co layer remains blocked).further increasing the applied magnetic field strength, the middle layermagnetic momentwill eventually rotate parallel to the applied field. Thefield strength required to do this is determined primarily by the strengthof the antiferromagnetic coupling of this layer to the bottom Co layer.The magnetization vs field strength of a typical Co/Ru/Co/Cu/Co ampleis shown n Figure 5. It behaves exactly as described above. Similar struc-tures have been used to measure oscillations in ferromagnetic interlayercoupling via Ru (24), Cu (74), and Pd (75).The dependenceof the resistance of the spin-engineered structure withfield shown n Figure 5 is unusual. The structure does indeed exhibit GMR,but as expected from the unusual variation of the magnetic structure withfield, the resistance of the structure increases for small fields. This is simplybecause the relative orientation of the two Co layers surrounding the Cu




n05/06/05.

Forpersonaluseonly.


13/34

GIANTMAGNETORESISTANCE 369

ield kOe)Figure 5 Schematic diagram of a spin-engineered structure in which two Co layers areferromagnetically exchange coupled through a Cu layer of the appropriate thickness with aninterlayer coupling strength J,* . The middle Co layer is antiferromagnetically coupled to athird (lower) Co layer through a thin Ru layer, with a coupling strength J . The moment ofthe lower Co layer is set to be equal to or slightly larger than the sum of the mom ents of thetop two Co layers. Th e m agnetic arrangement of the Co layers is shown for small, inter-mediate, an d strong magnetic fields left, middle and right, respectively). These arrangementscorrespond to the regions of the m agnetization vs field a nd corresponding resistance vs fieldcurves as indicated by the arrows. T he inset a t the right of the figure is an enlarged view ofthe low field region of the resistance vs field curve, which demonstrates an inverse GMReffect (see 24).

layer changes from being parallel to being antiparallel to one another.Thus this structure exhibits an inverse G M R effect, although its origin issimply related to the unusual field dependence of the magnetic con-figuration of the magnetic layers. At higher fields, the resistance of the




alreviews.org


Forpersonaluseonly

.


14/34

370 v~structure decreases as the magneticmomentsf all three magnetic ayersbecome arallel to the applied field, as shownn the figure.GIANT MAGNETORESISTANCE AT LOW FIELDS INMAGNETIC MULTILAYERSFor applications of giant magnetoresistance o magnetoresistive sensorsfor magneticrecording applications, large changes n resistance at lowfields of -~ 5 to 10 Oeare required (28). At present, the magnetoresistivecomponentf such sensors is comprised f a thin film alloy of Ni and Fe,permalloy (Py), NislFe~9. Permalloy exhibits a reasonably large mag-netoresistance at room emperatureof nearly 4% or fields of less than 1Oe for films of thickness> --- 1000~ (11). Note, however,hat the typicaloperating temperature in a high-capacity magneticdisk storage drive islikely to be significantly higher than room emperature,perhaps-~ 100C.An important stimulus to muchwork on GMRs to develop structureswith large GMRalues at the lowest possible fields. In addition to largelow-field GMR,he structures mustpossesscertain other properties if theyare to be useful as MR ensors for magneticdata storage applications.Theseproperties include magnetic tability and reproducibility; stabilityagainst the elevated temperatures required for processing of the MRdevice; low magnetostrictionso that the material will be insensitive tostrains inducedduring processing; stability against electromigration dueto the very high current densities (10v to l0g A/cm) the material willexperience in order to obtain the required signal; and stability againstcorrosion from the environmentwithin which he material will operate.Theseconditionsare quite stringent and difficult to meet.Dependence of Giant Magnetoresistance on Spacer LayerThicknessAt first glance, the very large GMRalues in Figure 1 are extremelypromising or MRead head applications. However,he saturation fieldsare clearly too high for useful application to MR ecording heads. Asshown n Figure 2, the magnitudeof the GMRecays rather slowly withincreasing Cu layer thickness in Co/Cumultilayers. Indeed, the GMRdecays approximately as the inverse of Cu layer thickness at low tem-peratures, where he mean-free path of the conduction electrons withinthe Cu layers is long compared o the Cu layer thickness. At highertemperatures, where he mean-freepath becomes omparable o the rangeof Cu layer thicknesses shown n Figure 2, the GMRecays more apidly,approximatelys AR/R c1/tcu exp~tc/~c~),where cu is the Cu ayer thick-ness and 2cu describes scattering within the Cu layer interior (67). This




n05/06/05.

Forpersonaluseonly.


15/34

GIANT MAGNETORESISTANCE 371functional dependencean be understood y considering wo effects. First,increasing he Cu layer thicknessdilutes the interfacial scattering regionsbecause he measuring urrent, which s parallel to the layers, is shuntedthrough the bulk of the Cu layers away from these regions. Second,scattering within he interior of the Cu ayers will clearly diminishhe flowof electrons betweenCo layers. This scattering can be described by ascattering length 2Cu,which s likely related to the mean-free ath in thickCu layers, where he contribution fromscattering at the surfaces of thefilm is small. Thus, for sufficiently thick Cu layers where mostof thecurrent is carried by the Cu layers, we can account for the variation ofGMR ith Cu layer thickness (67). In contrast, the interlayer exchangecoupling decays muchmore rapidly with increasing Cu layer thickness(22, 74, 76-79). For example,his meanshat for Cu ayers about twicethick as those contained in the Co/Cumultilayer in Figure 1, the GMRsonly reducedby a factor of two, but the saturation field is about 20 timessmaller. Thus GMRalues of almost 35% or fields of about 100 Oe arepossible for Cu layer thicknesses of _~ 20 ~ (77). Suchmaterials mayuseful for magnetic ield MR ensors.Giant Ma#netoresistance in Permalloy/Au MultilayersAs mentioned bove, oscillatory interlayer coupling has been observed na wide variety of transition metal multilayers. Whereashe oscillationperiod is similar for many ystems nd there is no simple variation of theperiod with spacer layer material, there is a systematic variation of theinterlayer coupling trength with the position of the transition metalspacerelement n the periodic table. Indeed, the coupling strength is found toincreaseexponentiallywith d-band illing across the 3d, 4d, and 5d periods.In addition, the couplingstrength increases systematically from he 5d to4d to 3d elements n any given column f the periodic table (25). Basedthese considerationsand aking into account hat noblemetal spacer layerstypically give the largest GMRalues, goodcandidates for the large GMRat low saturation fields (low interlayer coupling)are multilayers withspacer layers. Indeed, one of the first observations of GMRn magneticmultilayers was in perpendicularly magnetized nd magnetically weaklycoupled or uncoupledCo/Au/Coandwiches 80, 81). The two Co layersweresufficiently thin that the magnetic nterface anisotropywasable toovercome he demagnetization field from the Co magnetization so thatthe magneticmomentswere perpendicularly magnetized n zero magneticfield. Bypreparing he two Co ayers with slightly different thicknesses,and correspondinglydifferent magneticperpendicular anisotropies, themomentsf the two Co layers switchedat different field strengths. Thusthere was a certain field range where he Co moments ere aligned anti-


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly.


16/34

372 PARKINparallel to one another. The resistance of this structure was higher thanfor fields where the Co momentswere aligned parallel to one another. Thisstructure is similar to subsequent observations of GMRn permalloy/Cu/Co/Cumultilayers (82) and exchange-biased sandwiches (83) in whichsuccessive magnetic layers are designed to have different in-plane coerci-vities or switching fields, as discussed further below.

Antiferromagnetic coupling across Au spacer layers and oscillations inthe magnetic interlayer coupling with Au thickness have been observed inFe/Au/Fe sandwiches and Fe/Au multilayers (84-87), although only smallGMR alues were observed (86). Recently, oscillatory coupling viasandwiched between Co layers (88) and permalloy layers (89) hasfound. The latter system displays large MR alues at very low fields atroom temperature, as described below.The magnitude of the saturation magnetoresistance vs Au layer thick-ness is plotted in Figure 6a for three families of (111)-oriented Py/Aumultilayers. The multilayers were prepared by MBE eposition and weregrownon (0001) sapphire substrates. The [111] crystalline orientation wasestablished by using a thin Pt seed layer, ~20 ~ thick, grown at 600C.Related saturation fields determined from the magnetoresistance curvesare shown n Figure 6b. They correspond to the field at which the MR shalf that of the saturation MR.Four well-defined oscillations in saturationmagnetoresistance and saturation field are found as the Au layer thicknessis increased (indicated by AFn, n = 14 in Figure 6a). Typical resistancevs in-plane magnetic field curves are shown n Figure 7 for samples cor-responding to the four maxima n Figure 6a and b. Note that a smallamountof Au was deliberately inserted in the permalloy layers in two ofthe families of samples depicted in Figure 6, by coevaporation from thepermalloy and Au sources. This was an attempt to take advantage of thelikely surface-segregation of the Au from the Py layers during growth soas to reduce the possibility of magnetic pin-holes between the Py layers.Similar samples grown without Au in the Py layers show suppression ofthe MRpeak near tAu-~ 11 /~ (AFI). An analogous surfactant effecthas been reported for copper surface-segregation through cobalt duringepitaxy of exchange-coupledCo/Cu 11) multilayers (90). Note that alloy-ing the permalloy with Auhas little effect on the properties of the systemfor thicker Au layers, except that the resistivity of the multilayers isincreased. For example, at the second MR eak, the resistivity is increasedby about 30%, consistent with the decreased MR f these samples (seeFigure 6, open squares) relative to otherwise similar samples.Typical magnetization vs in-plane magnetic field are shown in Figure8a and b for two Py/Au multilayers from the AF1 and AF2 peaks. Themagnetization loops are consistent with the resistance curves shown n




n05/06/05.

Forpersonaluseonly.


17/34

12AFI

GIANT MAGNETORESISTANCE

0 I I I I0 10 20

~F2 =,a PYssAu4,/Au ~. Py/Au

/I I I I I30 40

373

1 O0 ~

o D

0 10 20 30 40Au hickness~)Figure Roomemperatureaturation a~qetoresistancea) and aturationield (b) vsspacer ayer hickness t 295K or threeseries of Py/Au ultilayers. he pen nd olid boxescorrespondo two amplesrown ithPYs~Au,4agneticayers,and he solidcirclescorrespondto samplesrepared ithPy ayers. Fourpeaksn MRnd aturation ield are observedn theAu hickness ange panned, s indicated y AF1,AF2,AF3, ndAF4from 9).Figure 7 and indicate antiferromagnetic coupling of the permalloy layers.The degree of antiferromagnetic coupling can be inferred from the rem-anent magnetization near zero field. Figure 8 demonstrates almost perfectantiferromagnetic alignment at the AF2peak but incomplete (_~ 45%)


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly.


18/34

37412

4

42

PARKIN

(o)

-400 -200 0 200 400

-40 -20 0 20 40Field

1281

6420-20Oe

b

-100 0 100

~F 4

I I I I I I I-10 0 10 20

Fi#ure 7 Room emperature resistance vs in-plane magnetic field curves for four samplescorresponding to the four maxima n MR hown in Figure 6, with Au layer thicknesses of(a) AFh 11.7 ,~, (b) AF2:21.5 /~, (c) AF3:29.8 /~, and (d) 41.0 ~. The correspondingresistivities of these samples at room temperature are -~43, 19, 13, and 10 ~tOhm-cm,respectively, in fields large enough o saturate the magnetization of the samples.

coupling at the AF 1 peak. The detailed relationship of magnetization andMR s explored in Figure 8c and d. Because permalloy is a soft magneticmaterial whose magnetization is saturated in small fields, the simplestmodel of GMRan be used in which the resistance is proportional to thecosine of the angle between the magnetic momentsof neighboring layers(91). Then the resistance of the structure is expected to vary as -M(H),where M(H) s the magnetization parallel to the applied field H. Figures8b and d show a comparison of resistance and -M~ vs field curves.Excellent agreement is obtained confirming the expected relationship. Notethat for the AF1 sample, the MR s compared with --(M--Mo) ~, whereM0 s the residual magnetization in small fields. M0 epresents the portionof the sample (-~ 55%) containing ferromagnetically coupled Py layers dueto structural defects. These layers do not contribute to the magneto-resistance. If there were complete AFcoupling of the Py layers, these datawould suggest that the magnitude of the MR t the AF1 peak would be


Annu

.Rev.Mater.Sci.1995.25:35738

8.Downloadedfromarjournals.annualreviews.org


Forpersonaluseo

nly.


19/34

GIANT MAGNETORESISTANCE 375

-,~oo-2oo 20o ,~ooFieldOe)

8

--100 --50 0 50Field Oe)

6

lOO

Figure 8 Magnetization vs field curves, (a) and (b), for two Py-Au/Aumultilayers fromAFI and AF2 peaks shown in Figure 6. In (c) and (d), corresponding resistance vs fieldcurves (solid lines) and plots of -(M-Mo)vs field (solid circles) are shown. M0 s theremanent magnetization.

much higher (MR ~ 20%) for multilayers containing Py-Au magneticlayers and yet higher still for pure Py layers.The magnetic interlayer coupling in the Py/Aumultilayers is relativelysmall. For example,at the AF1 peak, JAy is ~- 0.02 erg/cm2. This is approxi-mately five times larger than that found in sputtered (111) textured Py/Aumultilayers (92). Although the magnitude s similar to that found in sput-tered Py/Cu multilayers (93), if we assume JAF is proportional to 2,the magnetic coupling through epitaxial (111) Au is about an ordermagnitude smaller than the coupling found at AF1 in epitaxial (111)(94). Note that as mentionedabove for Co/Cumultilayers, the strengththe coupling falls off rapidly with thicker Au layers. Indeed, as shown nFigure 6b, the coupling varies as approximately 1/t Au, where n = -~3.2.Such a dependence is more rapid than predicted in simple Ruderman-Kittel-Kasuya-Yosida (RKKY)models of the coupling (95).Interlayer coupling is usually found to be weaker in polycrystallinemultilayers than in similar single crystalline multilayers. Note that thelowest reported saturation fields at room temperature in magnetic multi-layers are found in sputtered Py/Aumultilayers where changes in resistanceof ~ 1 to 2%/Oecan be obtained (92).




alreviews.org


Forpersonaluseonly

.


20/34

Giant Magnetoresistance in Exchange-Biased SandwichesAn mportant structure for obtaining GMRt low magnetic ields is shownin Figure 9. This structure, an exchange-biasedandwich EBS), s in itssimplest embodiment, F~/S/F~/FeMn. The EBScontains two ferro-magnetic ayers, FI and FI~, and a single nonmagneticpacer layer, S, in

Field

FreeFerromagnetSpacerLayer

PinnedFerromagnetAntiferromagnet

Figure 9 Schematic diagram of an exchange-biased sandwich structure.


Annu.Rev.

Mater.Sci.1995.25:357388.Downloadedfromarjournals.annu

alreviews.org

byMax-Planck-Gesellschafton

05/06/05.

Forpersonaluseonly.


21/34

GIANT MAGNETORESISTANCE 377which one of the magnetic layers, F~, is exchange coupled to an anti-ferromagnetic layer. EBSstructures with antiferromagnetic exchange biaslayers of FeMn 83), NiO (96), and TbFe (97) have been reported.structure takes advantage of a phenomenonirst discovered more than 30years ago in oxidized Co particles (98) and subsequently extensively studiedin a numberof thin film systems (99-101). This phenomenon, ften referredto as exchange anisotropy, arises from an interfacial magnetic exchangecoupling between an antiferromagnetic layer and a ferromagnetic layer.Under appropriate conditions, the exchange anisotropy results in a unidi-rectional anisotropy in the F layer such that its magnetic hysteresis loopis centered about a non-zero field, a bias field Ha. In contrast, providingthat the magnetic coupling of F~ and F~ via the spacer layer is sufficientlyweak, the magnetichysteresis loop of F~ is centered close to zero field. Themoments f F~ and FII are thus aligned antiparallel for somefield rangeintermediate between zero and Ha. A resistance vs field curve is shown nFigure 10a for a typical EBS tructure, where F~ and Eli are Py and S isCu. The current and field are aligned along the unidirectional anisotropydirection. The structure displays a giant MR ffect exactly analogous tothat in multilayers with a higher resistance for fields where F~ and F~ areantiparallel. As is found for multilayered structures (22, 93), replacing thePy layers with Co layers of the same thickness increases the magneto-resistance of the structure by approximately a factor of two, as shown nFigure 10b.The quest for structures exhibiting large changes in resistance per unitfield has motivated a great deal of work in GMR. he simplest structurewith the largest change in resistance per unit field at room emperature isshown n Figure 10c. This structure is comprised of elements of the twoEBSstructures shown in Figure 10a and b. As can be seen from Figure10a, the Py/Cu/Py EBShas a small switching field and value~ of AR/Rofabout 2-4% at room temperature. Similar Co/Cu/Cu EBS have higherMR alues, as high as -~9% at room temperature, but larger switchingfields. The increased switching field results from the higher magnetic ani-sotropy of Co as compared with Py. To obtain the highest values of MRper unit field, EBS andwiches such as those shown in Figure 10c havebeen developed (102). These structures are essentially comprised ofPy/Cu/Py EBS in which thin Co layers are introduced at each Py/Cuinterface. The Co layers need only be approximately 1 atomic layerthick, or just sufficient to completely cover the Py/Cu interface. Thesestructures also dramatically demonstrate the overwhelming mportance ofspin-dependent scattering at the magnetic/nonmagnetic interfaces (102,103).




alreviews.org


Forpersonaluseonly

.


22/34

378 PARKIN

FeMnNiFeCuNiFe

0Field

FeMn~4Cu z:

o

oReid (0~6

~FeMnNiFe o~ ~--Co ~4NIFe 2

0Flelcl

~/C@C@CW~t=3A

Figure 10 Room emperature resistance vs field curves for (a) Si/Py(53/~)/Cu(32/~)/Py(22/~/FeMn(90/~)/Cu(10~); (b) the same structure with the Py (permalloy = Nis~Fe~9)replaced by Co; and (c) the same structure as in (a) but with 3.0 k-thick Co layers addedeach Py/Cu interface. (Note the thicknesses of the Py layers have correspondingly beenreduced by 3.0/~..)


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly.


23/34

GIANT MAGNETORESISTANCE 379DEPENDENCE OF GIANT MAGNETORESISTANCEON STRUCTUREIn multilayers, the magnitudeof the saturation giant magnetoresistance isstrongly dependenton the details of the multilayer structure. In particular,the GMR ill depend on the thicknesses of both the magnetic and non-magnetic layers and the numberof bilayers, as well as on the compositionand thickness of any seed layers and capping layers. In addition, the GMRis sensitive to the detailed morphology of the structure, which can bechanged by varying deposition conditions such as the deposition tempera-ture, the detailed method of preparation, and the substrate. The GMRwill also be decreased if the magnetic layers are not completely coupledantiferromagnetically to one another in the high resistance state. Thismeans hat it is very easy to obtain different values of GMRor nominallythe same structure, and it is very difficult to makemeaningfulcomparisonsof results on individual samplesfrom different groups. It is most useful tostudy sets of samples made under identical conditions where one aspect ofthe structure is systematically varied.Dependence of GMR on Number of BilayersPerhaps one of the most obvious aspects of the structure of a magneticmultilayer that can be varied is the overall thickness of the multilayer, asdetermined by the number of bilayers N. The magnitude of the GMRanbe, and usually is, strongly dependent on N. Figure 11 shows typicalresistance vs field curves for representative samples from two series ofFe/Cr multilayers with different N. The series differ only in the thicknessesof the Cr overlayer to and the Cr underlayer tu. A minimum r underlayerthickness of --- 10/~ was required for the growthof well-layered structures.As can be seen from Figure 11, the measured GMRs increased withincreasing N in both series of structures. However, he rate of increase ofGMR ith N is significantly different for the two series. This is clearlydemonstrated in Figure 12, where the detailed dependence of MR n N isdisplayed for the two families of structures. The magnetoresistance isobserved to increase muchmore rapidly with N for the [Fe/Cr]N multilayerswith very thin Cr under- and overlayers than for the structures with thickerCr layers. The origin of this effect is similar to that described above forthe dependence of GMR n Cu layer thickness in Co/Cu multilayers. TheGMRn the Fe/Cr multilayers is reduced in proportion to the measuringcurrent passing through the Cr under- and overlayers. The dilution of the[Fe/Cr]N portion of the multilayer contributing to the giant MR,by thenoncontributing portion of the structure, will be decreased as the relativeproportion of the noncontributing portion decreases with increased N.


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly.


24/34

380 PARKIN

30

20

A

N= 42

-12 -8 -4 0 4 8 12Field (kOe)

3o

2o

B

10

o-12 -8 I I I0Field kOe)

4 8 12

FixTure 11 Resistance vs field curves at 4.2 K for several membersof two families ofFe/Cr multilayers with different numberof bilayers. (a) This correspmads to data for Fe/Crmultilayers of the form Si/10/~ Cr/[18 .~ Fe/9 .~. Crib/10 ~ Cr with N ~ 2, 4, I0, and 42. (b)This corresponds to data for Si/115 ,~ Cr/[16/~ Fe/l 1.5/~ Cr]~/115 ~ Cr with N = 3, 24,and 50.


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly.


25/34

50l I i

GIANT MAGNETORES1STANCE 381I I I I I I

10~ CFunder (and over layers

10 ~15~ run er an ove layer

0 u I I I I I I I I I I0 10 20 30 40 50NFigurel2Dependencef saturationmagnetoresistancennumberf bilayers or two eriesof Fe/Crmultilayers ith tructureof the formupperine) Si/10/~Cr/[18~ Fe/9 ~ Cr]N/10,~. Cr; and lowerine)Si/115/~ r/[16/~ e/l 1.5 ~. Cr]N/115/~.r.

The results shown n Figure 12 are rather similar to those found formanymagnetic multilayered systems. It is not surprising, therefore, thatthe largest GMR alues in Fe/Cr (and other multilayers) are foundmultilayers with large values of N (33, 34). The dependence on N is notcompletely determined by a simple dilution effect; there are several otherfactors. First, an important and very simple factor affecting the dependenceof GMRn Nis that the two magnetic layers at either end of the multilayercontribute to GMR nly half as muchas the magnetic layers within theinterior of the multilayer (assumingdiffuse scattering at either extremityof the multilayer). Second, as described in the very first models of GMR(59, 61, 62), the GMR ill be increased if the electrons propagate acrossmanymagnetic/nonmagnetic interfaces within a conduction carrier mean-free path. WhenN is small compared with some mean-free path withinthe multilayer, the GMRs reduced. Thus it follows that the dependenceofGMRn N will be temperature dependent, as described by the temperaturedependence of the mean-free path. This effect is muchmore important asthe conductivity of the multilayer increases: for example, in multilayersystems containing metals with long mean-free paths such as Cu, Ag,and Au and in single-crystal, defect-free multilayer systems. Third,




alreviews.org


Forpersonaluseonly

.


26/34

382 PARKINthe resistance of the film will increase as the film thickness is decreasedbecauseof scattering at the film surfaces. This leads to a decreasedGMReffect.Finally, it mustbe noted that the structure of the multilayer tself willvary with total film thickness.For polycrystalline ilms, the grain size oftenincreases in proportion o the film thickness. TheGMRs clearly sensitiveto the grain size (35). Thevariation in number f grain boundarieswithfilm thickness will lead to variations in film resistivity and hus GMR.orsingle-crystal films with no large-angle grain boundaries, the changes nstructure with film thickness maybe moresubtle. For Fe/Cr, it has beenproposed that increases in GMR ith N maybe attributed to an increasein the roughness f the Fe/Cr nterfaces at someength scale (104). Indeed,a particularly important aspect of the origin of GMR, hich s not dis-cussed n detail in this article, is the role of the structure of the interfaces,which emains controversial.Dependence of GMR.on Magnetic Layer ThicknessA basic assumption of the resistor network model of GMRescribedabove s that the spin-dependent cattering giving rise to the GMRri-ginates purely within the interior of the magnetic ayers, i.e. frombulkspin-dependent cattering. This model an be readily generalized to allowfor spin-dependentnterfacial scattering by addingadditional resistors inthe network epresenting the scattering in the interfacial regions. Therelative contribution of spin-dependent cattering from bulk scatteringand fromspin-dependent cattering at tlae interfaces between he magneticand spacer ayers is a subject of great current interest.A simple experiment o examine he role of interfacial vs bulk spin-dependent cattering is to vary the interfacial content of the magneticmultilayer by varying he thickness of the magnetic ayers. Figure 13 showsvarious resistance vs field curves measured t 4,2 K for sets of otherwiseidentical Fe/Crmultilayers n which he Fe layer thickness s systematicallyvaried from -~ 1.5 to 300 ,~. The magnitudeof the saturation GMRorthese samples s plotted in Figure 14 vs Fe layer thickness. These datashow hat maximum MRalues are obtained for thin Fe layers -~ 8 ~-thick, which tronglysuggests hat interfacial scattering is of predominantimportance. Note, however, that the GMRs diluted by the presence ofnecessary underlayers(for growth) and overlayers (for protection againstcorrosion). This means hat the Fe thickness for which maximum MRis obtained will be affected by the presence of these layers and by thethickness of the Cr spacer layers themselves. ncreasing the thickness ofthese layers will increase the thickness of Fe necessary o obtain maximum




n05/06/05.

Forpersonaluseonly.


27/34

60

40

20

0 0 20 40

GIANT MAGNETORESISTANCEc~(eA)[vo(tr,)/c~(gA))=o

~ -,, tr.(~)

LV;, ,oI-,\ \:, ,\I ~% "~ N 17.5

383

60 0 10 20H kOe)

30 40 50

Figure 13 Resistance vs field curves for several Si/Cr(t,)/[Fe(tzo)/9 Cr]2o mutilayers wi thvarying Fe layer thickness tF. The Cr underlayer thickness tc, is ~- 10/~. On he left panel,the sample with tFo = 1.5 A has the smaller GMR.On the right panel, the saturation fieldsystematically increases from the largest to the smallest value of t~ shown.

60

40

20

I I I I I T1 O0 200 300Fe layer thickness

Figure 14 Dependenceof saturation magnetoresistance at 4.2 K vs Fe layer thickness forstructures Si/9 ~ Cr/[Fe(tvo)/9 ,~ Cr]~o, for Fe layer thicknesses ranging from 1.5 to 300/~.For thick Fe layers, the anisotropic magnetoresistance (AMR) ecomesa significant fractionof the GMR ffect. For these samples, shown as filled circles, the AMRwas measured anda correction applied.




alreviews.org


Forpersonaluseonly

.


28/34

384 PARKINGMR.Thus it is not straightforward to interpret such studies withoutappealing to detailed theoretical models. Note, however, that GMRmodelspredict a maximumn the GMRffect for Fe layers significantly thickerthan those found in Figure 14 for any significant spin-dependent bulkscattering (105). Finally, note that the MR ecreases approximately as theinverse Fe layer thickness for Fe layers, ranging in thickness from 20 to300/~, consistent with dilution of the interracial scattering regions, andpredominant interfacial spin-dependent scattering. Similarly, in granularalloys, the magnitude of GMRecreases approximately as the inverse sizeof the magnetic particles, as shown n Figure 15.

3O

2O

10

Anneal emp(oc)

35o~.~~I55o ..:;,:...I I I I

-10 0F:ield (Oe)

H perpendicularmeasuredt 4.2K

10

Figure 15 Variation of GMRn a single crystal (11 I) Co28Cu72 ranular alloy at 4.2 K withannealing temperature of the alloy. The granular alloy was deposited at 200C and suc-cessively annealed for 1 h at 250, 350, 450, and 500C. The size of the Co particles asdetermined from small-angle grazing incidence X-ray scattering is shown by the size of arepresentative particle. Note that the largest particle size for the sample annealed at 550Cis "- 135/~ (see 115 for moredetails).




n05/06/05.

Forpersonaluseonly.


29/34

GIANT MAGNETORES1STANCE 385CONCLUSIONSInterest in magnetic nano-structured metals has substantially increased inrecent years with the discovery of giant magnetoresistance and oscillatoryinterlayer exchange coupling in magnetic multilayers and related struc-tures. This brief review has discussed only a small amountof the publishedwork in this field, which continues to grow at an enormouspace. Importantomissions include work on detailed models of interlayer coupling andGMR;magnetotransport in magnetic multilayers and sandwiches in whichthe current flows perpendicular to the layers (31, 32, 106-110); the detailedrole of the influence of the structural morphology, specially of the inter-faces, on GMR;detailed discussions of magnetic granular alloys andmultilayered granular structures (42); spin-dependent quantum con-finement of electrons in the magnetic potentials created by the mag-netic/nonmagnetic interfaces; and the influence of such potentials on theinterlayer coupling and GMRsee, for example, PD Johnson, this volume).Magnetic nano-structures display other anomalous transport propertiesincluding giant magneto-thermopower and giant magneto-thermalconductivities (see, for example, 111), which have not been discussedherein.It seems quite clear that giant magnetoresistance is intimately relatedto scattering at the interfaces between the magnetic and nonmagneticcomponents n both magnetic-layered structures and granular alloys. Thelargest GMRalues are usually found for structures in which the inter-facial component s maximized, by reducing either the thickness of mag-netic layers or the size of the magnetic particles. Similarly, the GMRis extremely sensitive to the detailed chemical nature of the interfaces.Modifications of these interfaces by dusting with just sub-monolayer tomonolayer equivalent coverages of impurity elements results in dramaticvariations in GMR alues (102). Indeed, GMRmay thus be used to probethe electronic character of the magnetic/nonmagnetic nterfaces, just asoscillatory interlayer coupling can be used to probe details of the Fermlsurface topology of the spacer layer material (38, 52) and perhaps themagnetic layer material (112-114). The ability to tailor-make magneticnano-structured materials with properties especially engineered for par-ticular applications is very attractive and is likely to lead to importantapplications for such structures in the near future.ACKNOWLEDGMENTSI would like to thank manycolleagues, especially Drs. R F C Farrow, RMarks, A R Modakand Prof. D J Smith for their contributions to parts




alreviews.org


Forpersonaluseonly

.


30/34

386 PARKINof the workdescribed in this paper. I amespecially grateful to KevinRoche or technical support.

Any nnualReviewchapter as well as any article cited in an nnualReviewchaptermayhe purchased rom the AnnualReviewsPreprints and Reprints service.1-800-347-8007; 415-259-5017; emaih [email protected]

Literature Cited1. Schuller IK, Falco CM. 1982. Pro9.Phys. 4:4922. Schuller IK. 1988. In Physics, Fabri-cation, and Applications of MultilayeredStructures, ed. P Dhez, C Weisbuch,p.139. NewYork: Plenum3. Falicov LM, Pierce DT, Bader SD,Gronsky R, Hathaway KB, et al. 1990.J. Mater. Res. 5:12994. Mathon J. 1991. J. Magn. Magn. Mat-ter 100:5275. Fert A, BrunoP. 1994. See Ref. 116, p.826. Hathaway KB. 1994. See Ref. 116, p.457. Heinrich B, Cochran JF. 1993. Adv.Phys. 42:5238. Levy PM. 1994. In Solid State Physics,ed. H Ehrenreich, D Turnbull, 47: 367.

New York: Academic9. Parkin SSP. 1994. See Ref. 116, p. 14810. Pierce DT, Unguris J, Celotta RJ. 1994.See Ref. 116, p. 11711. McGuire TR, Potter RI. 1975. IEEETrans. Magn. MAG-11:101812. Binasch G, Grfinberg P, Saurenbach F,Zinn W. 1989. Phys. Rev. B 39:482813. Baibich MN, Broto JM, Fert A,Nguyen van Dau F, Petroff F, et al.1988. Phys. Rev. Lett. 61:247214. Gr~nberg P, Schreiber R, Pang Y,Brodsky MB, Sowers H. 1986. Phys.Rev. Lett. 57:244215. Carbone C, Alvarado SF. 1987. Phys.Rev. B 36:244316. Joyce BA. 1985. Rep. Prog. Phys. 48:163717. Ploog K. 1988. Angew. Chem. Int. Ed.Engl. 27:593118. Farrow RFC, Marks RF, Harp GR,Weller D, Rabedeau TA, et al. 1993.Mater. Sci. Eng. Rep. Rll: 15519. Parkin SSP, More N, Roche KP. 1990.Phys. Rev. Lett. 64:230420. Parkin SSP, York BR. 1993. Appl.Phys. Lett. 62:184221. Parkin SSP, Li ZG, Smith DJ. 1991.Appl. Phys. Lett. 58:271022. Parkin SSP, Bhadra R, Roche KP.1991. Phys. Rev. Lett. 66:2152

23. Kano H, Kagawa K, Suzuki A, OkabeA, Hayashi K, Aso K. 1993. Appl.Phys. Lett. 63:283924. Parkin SSP, Mauri D. 1991. Phys. Rev.B. 44:713125. Parkin SSP. 1991. Phys. Rev. Lett. 67:359826. Kryder MH. 1993. Annu. Rev. Mater.Sci. 23:41127. Grochowski E, Thompson DA. 1994.IEEE Trans. Magn. 30:379728. Ciureanu P. 1992. In Thin Film Resis-tive Sensors, ed. P Ciureanu, S Mid-delhoek, p. 253. Bristol: lOP Pub.29. Daughton JM. 1992. Thin Solid Films216:16230. Prinz G. 1990. Science 250:109231. Johnson M. 1993. Science 260:32032. Johnson M. 1993. Appl. Phys. Lett. 63:143533. Fullerton EE, Conover MJ, MattsonJE, Sowers CH, Bader SD. 1993. AppLPhys. Lett. 63:169834. Schad R, Potter CD, Belien P, Ver-banck G, Moshchalkov VV, Bruyn-seraede Y. 1994. Appl. Phys. Lett. 64:350035. Modak AR, Smith DJ, Parkin SSP.1994. Phys. Rev. B, 50:423236. Parkin SSP, Roche KP, Suzuki T. 1992.Jpn. J. Appl. Phys. 31:L1246

37. Johnson MT,Coehoorn R, de Vries J J,McGeeNW,aan de Stegge J, BloemenPJ. 1992. Phys. Rev. Lett. 69:96938. Bruno P, Chappert C. 1992. Phys. Rev./3 46:26139. Berkowitz AE, Mitchell JR, Carey MJ,Young AP, Zhang S, et al. 1992. Phys.Rev. Lett. 68:374540. Xiao JQ, Jiang JS, Chien CL. 1992.Phys. Rev. Lett. 68:374941. Parkin SSP, Farrow RF, RabedeauTA, Marks RF Harp GR, et al. 1993.Eur. Phys. Lett. 22:45542. Hylton TL, Coffey KR, Parker MA,Howard JK. 1993. Science 261:102143. Samant MG, Strr J, Parkin SSP, HeldGA, HermsmeierBD, et al. 1994. Phys.Rev. Lett. 72:1112


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly.


31/34

44. Unguris J, Celotta RJ, Pierce DT. 1992.Phys. Rev. Lett. 69:112545. Pappas DP, Prinz GA, Ketchen MB.1994. Appl. Phys. Lett. 65:340146. Unguris J, Celotta R J, Pierce DT. 1991.Phys. Rev. Lett. 67:14047. Purcell ST, Folkerts W, Johnson MT,McGeeNW, ager K, et al. 1991. Phys.Rev. Lett. 67:90348. Demokritov SO, Tsymbal E. 1994. J.Phys. Conden. Matter 6:714549. CochranJF. 1994. See Ref. 116, p. 22350. Hillebrands B, Gtintherodt G. 1994.See Ref. 116, p. 25851. Krebs JJ, Lubitz P, Chaiken A, PrinzGA. 1989. Phys. Rev. Lett. 63:164552. Edwards DM, Mathon J, Muniz RB,Phan MS. 1991. Phys. Rev. Lett. 67:49353. Mathon J. 1991. Contemp. Phys. 32:14354. Stiles MD.1993. Phys. Rev. B48:723855. Mort NF. 1964. Adv. Phys. 13:32556. Campbell IA, Fert A. 1982. In Ferro-magnetic Materials, ed. EP Wohlfarth,3: 747. Amsterdam: North-Holland57. Dorleijn JW. 1976. Philips Res. Rep.31:28758. Rossiter PL. 1987. The Electrical Resis-tivity of Metals and Alloys, Cambridge:Cambridge Univ. Press59. Camley RE, Barnas J. 1989. Phys. Rev.Lett. 63:66460. Levy PM, Ounadjela K, Zhang S,Wang Y, Sommers CB, Fert A. 1990.J. Appl. Phys. 67:591461. Barnas J, Fuss A, Camley RE, Gr~in-berg P, Zinn W. 1990. Phys. Rev. B 42:811062. Barthelemy A, Fert A. 1991. Phys. Rev.B43:1312463. Inoue J, Oguri A, Maekawa . 1991. J.Phys. Soc. Jpn. 60:37664. Edwards DM, Muniz RB, Mathon J.

1991. IEEE. Trans. Ma9. 27: 354865. Trigui F, Velu E, Dupas C. 1991. J.Magn. Magn. Matter 93:42166. Zhang S, Levy PM, Fert A. 1992. Phys.Rev. B. 45:868967. Parkin SSP, Modak AR, Smith DJ.1993. Phys. Rev. B 47:913668. Barthelemy A, Fert A, Baibich MN,Hadjoudj S, Petroff F, et al. 1990. J.Appl. Phys. 67:590869. Dieny B, Gavigan JP, Rebouillat JP.1990. J. Phys. Condens. Matter 2:15970. Dieny B, Gavigan JP. 1990. J. Phys.Condens. Matter 2:18771. Folkerts W. 1991. J. Magn. Magn.Matter 94:30272. Parkin SSP, Mansour A, Felcher GP.1991. Appl. Phys. Lett. 58:147373. Heinrich B. 1994. See Ref. 116, p. 195

GIANT MAGNETORESISTANCE 38774. Bloemen PJH, van Dalen R, de JongeWJM, Johnson MT, aan de Stegge J.1993. J. Appl. Phys. 73:597275. Bloemen PJH, de Jonge WJM, Don-kersloot HC. 1993. J. Appl. Phys. 73:452276. Cebollada A, Miranda R, SchneiderCM, Schuster P, Kirschner J. 1991. J.Magn. Magn. Matter 102:2577. Parkin SSP. 1992. In MagneticSurfaces, Thin Films and Multilayers,ed. SSP Parkin, H Hopster, J-PRenard, T Shinjo, WZinn, 231: 211.Pittsburgh: Mater. Res. Soc. Symp.Proc.78. Johnson MT, Purcell ST, McGee NW,Coehoorn R, aan de Stegge J, HovingW. 1992. Phys. Rev. Lett. 68:268879. Qiu ZQ, Pearson J, Bader SD. 1992.

Phys. Rev. B 46:865980. Velu E, Dupas C, Renard D, RenardJP, Selden J. 1988. Phys. Rev. B 37:66881. Dupas C, Beauvillain P, Chappert C,Renard JP, Trigui F, et al. 1990. J.Appl. Phys. 67:568082. Shinjo T, Yamamoto . 1990. J. Phys.Soc. Jpn. 59:306183. Dieny B, Speriosu VS, Parkin SSP,Gurney BA, Wilhoit DR, Mauri D.1991. Phys. Rev. B. 43:129784. Fuss A, Demokritov S, Grt~nberg P,Zinn W. 1992. J. Magn. Magn. Matter103:L22185. Marty A, Gilles B, Eymery J, Cham-berod A, Joud JC. 1993. J. Magn.Magn. Matter 121:5786. Shintaku K, Hosoito N, Shinjo T.1993. J. Magn. Magn. Matter 121:41387. Sato H, Kobayashi Y, Aoki Y, Shin-taku K, Hosoito N, Shinjo T. 1993. J.Phys. Soc. Jpn. 62:338088. Grolier V, Renard D, Bartenlian B,Beauvillain P, Chappert C, Dupas C,Ferre J, Galtier M, Kolb E, Mulloy M,Renard JP, Veillet P. 1993. Phys. Rev.Lett. 71:302389. Parkin SSP, Farrow RFC, Marks RF,Cebollada A, Harp GR, Savoy R. 1994.Phys. Phys. Lett. 72:371890. Harp GR, Parkin SSP, Farrow RFC,Marks RF, Toney MF, Lam QH,Rabedeau TA, Savoy RJ. 1993. Phys.Rev. B 42:872191. Chaiken A, Prinz GA, Krebs JJ. 1990.J. Appl. Phys. 67:489292. Parkin SSP, Rabedeau TA, FarrowRFC, Marks R. 1994. J. Appl. Phys.76:661793. Parkin SSP. 1992. Appl. Phys. Lett. 60:51294. Parkin SSP, Marks RF, Farrow RFC,Harp GR. Lam QH, Savoy RJ. 1992.Phys. Rev. B 46:9262


Annu.Rev.

Mater.Sci.1995.25:357388.Downloadedfromarjournals.annu

alreviews.org

byMax-Planck-Gesellschafton

05/06/05.

Forpersonaluseonly

.


32/34

388 PARK1N95. Roth LM, Zeiger H J, Kaplan TA.1966. Phys. Rev. 149:51996. AnthonyTC, BrugJA, Zhang,S. 1994.IEEETrans. Mag.30:381997. Freitas PP, Leal JL, MeloLV,OliveiraNJ, Rodrigues L, Sousa AT. 1994.

AppLPhys. Lett. 65:40398. MeiklejohnWH,BeanCP. 1959. Phys.Rev. B 102:141399. Yelon A. 1971. In Physics of ThinFilms: Advancesn Research nd Devel-opment, ed. MFrancombe, R. Hoft-man, 6: 205. NewYork: Academic100.TsangC, Lee K. 1982. J. Appl. Phys.53:2605101.Parkin SSP,Deline V, Hilleke R, Fel-cher GP. 1990. Phys. Rev. B 42:10583102.Parkin SSP.1993.Phys. Rev. Lett. 71:1641103.ParkinSSP.1992.Appl. Phys. Left. 61:1358104. Fullerton EE, Kelly DM,GuimpelJ,Schuller IK, Bruynseraede Y. 1992.Phys. Rev. Lett. 68:859105. LevyPM,ZhangS, Fert A. 1990.Phys.Rev. Lett. 65:1643106. Pratt WP, Lee SF, Slaughter JM,LoloeeR, SchroederPA, BassJ. 1991.Phys. Rev. Lett. 66:3060

107. Lee SF, Pratt WP,LoloeeR, SchroederPA, Bass J. 1993. Phys. Rev. B 46:548108. YangQ, HolodyP, Lee S-F, HenryLL,LoloeeR, SchroederPA, et al. 1994.Phys. Rev. Lett. 72:3274109.Gijs MA,Giesbers JB, Lenczowski K,JanssenHH. 993.Appl. Phys. Lett. 63:111110. Gijs MA,Lenczowski K, van de Veer-donk R J, Giesbers JB, Johnson MT,aan de Stegge JB. 1994. Phys. Rev. B50:16733111. Shi J. 1994.Magneto-transportroper-ties of giant magnetoresistiveystems.PhD hesis. Univ. Illinois, Urbana112. BrunoP. 1993.Eur. Phys. Lett. 23:615113. Bloemen PJH, Johnson MT, van deVorstMT,Coehoorn , de Vries J J, etal. 1994.Phys. Rev. Lett. 72:764114. Okuno SN, Inomata K. 1994. Phys.Rev. Lett. 72:1553115. Rabedeau TA, Toney M, Marks RF,Parkin SSP, Farrow RFC, Harp G.1993. Phys. Rev. B48:16810116.Heinrich B, Bland JAC, eds. 1994.Ultrathin Magnetic tructures, Vol. 2.Berlin: Springer-Verlag


Annu.Rev

.Mater.Sci.1995.25:357388.D


alreviews.org


Forpersonaluseonly.


33/34


ownloadedfromarjournals.annualreviews.org


n05/06/05.

Forpersonaluseonly.


34/34


ownloadedfromarjournals.annualreviews.org


n05/06/05.

Forpersonaluseonly.

parkin annurev

Documents