Arrays of nanowires of magnetic metals and multilayers: Perpendicular GMR and magnetic properties

N ELSEVIER Journal of Magnetism and Magnetic Materials 175 (1997) 127 136

~ H Journal of magnetism and magnetic

~ I ~ materials

Arrays of nanowires of magnetic metals and multilayers: Perpendicular GMR and magnetic properties

L. P i r a u x a'*, S. D u b o i s a, J .L. D u v a i P , K. O u n a d j e l a b, A. F e r t c

UnitO de Physico-Chimie et de Physique des MatOriaux, UniversitO Catholique de Louvain, B-1348 Louvain-la-Neut~e, Belgium b lnstitut de Physique et Chimie des Mat~riaux de Strasbourg, F-67037 Strasbourg, Franee

c Unit~; Mixte de Recherche du Centre National de la Recherche Scient!fique et de Thomson, Laboratoire Central de Recherches Thomson, 91404 Orsay, France et Universit~ Paris-Sud, Bat. 510, 91405 Orsay, France

Abstract

The template strategy combined with electrodeposition techniques have been used to fabricate arrays of nanowires of magnetic metals and multilayers in the cylindrical pores of track-etched polymer membranes. The giant magnetoresis- tance effects have been investigated in two different types of multilayered nanowires systems: Co/Cu and NisoFe2o/Cu. In addition, a comparative study of the magnetic properties of sub-micron Ni, Co, Fe and NisoFe2o wires is made by means of anisotropic magnetoresistance and magnetization experiments.

PACS: 72.15.Gd; 72.10.Fk; 75.50.Rr

Keywords: Nanowires; Giant magnetoresistance; Multilayers; Magnetoresistance perpendicular

1. Introduction

Since the pioneering work of Pratt et al. [1], measurements of the giant magnetoresistance (GMR) of magnetic multilayers in the CPP (current perpendicular to the planes) geometry have become increasingly attractive. The growth of magnetic multilayered nanowires in the cylindrical pores of track-etched polymer membranes enables to inves- tigate GMR effects in the perpendicular geometry

*Corresponding author. Fax: + 32 10 473452; e-maih [email protected].

and provides a valuable testing ground for theories describing the CPP-GMR in various limits [2-6]. In this paper, a review is given of recent GMR data in two types of multilayered nanowires systems: Co/Cu and NisoFe2o/Cu. The variation of the CPP-GMR of Co/Cu multilayered nanowires has been explored with the thicknesses of the Co and Cu layers varying over very wide ranges (between the nanometer and the micrometer ranges). Com- bining the data obtained in the low-temperature range (where spin-mixing effects are negligible) with theoretical predictions of the Valet Fert (VF) model [7] led to an estimation of the spin diffusion lengths (SDL) in the nonmagnetic and

0304-8853/97/$17.00 ~C 1997 Elsevier Science B.V. All rights reserved PII S 0 3 0 4 - 8 8 5 3 ( 9 7 ) 0 0 1 5 7- 1

128 L. Piraux et al. /Journal of Magnetism and Magnetic Materials 175 (1997) 127 136

ferromagnetic layers. For applications, large resis- tance changes at low fields for temperatures at or above room temperature are required. Thus, multi- layers containing magnetically soft NisoFe20 layers are of particular interest. Two types of structures have been studied: conventional NisoFe2o/Cu multilayers and multilayers composed of NisoFe2o/ Cu/NisoFe2o trilayers magnetically isolated by long Cu rods. Finally, electrodeposited nanowires are of great interest because they provide a relative- ly simple and inexpensive way to study the mag- netic properties of nanoscale objects [8]. Due to the shape anisotropy, those elements are predomi- nantly magnetized along their length. However, there are still open questions about the mechanisms responsible for the magnetization reversal. We have recently shown that the Ni and Co-based systems exhibit different magnetic and magnetoresistive be- haviors due to competing crystal anisotropy in the Co-based system [9, 10]. In this paper, we also report on preliminary data obtained on arrays of Fe and NisoFe2o sub-micron wires.

In Section 2, we briefly report on the preparation of arrays of nanowires of magnetic metals and multilayers. The GMR properties of these multi- layered nanowires are discussed in Section 3. Finally, a comparative study of anisotropic mag- netoresistance and magnetic properties in arrays of sub-micron magnetic wires is made in Section 4.

2. Growth of arrays of electrodeposited nanowires of magnetic metals and multilayers

Track-etched polycarbonate membranes were used as nanoporous host material for the growth of multilayered nanowires [11]. A gold film serving as cathode was first evaporated on one side of the membrane. The membrane sample is then placed in a home-made teflon cell and a 0.1 cm 2 area is exposed to the electrolyte. Electrodeposition is per- formed using an EG&G Princeton Applied Re- search Model 283 potentiostat/galvanostat under quiescent conditions at T ~ 25°C. Ferromagnetic Co, Ni, Fe and NisoFe2o nanowires were elec- trodeposited in nanoporous membranes with dia- meters ranging from 30 to 500 nm using a sulphate bath. Electrodeposited Co/Cu and NiFe/Cu multi-

layered nanowires were made from a single bath using a pulsed deposition technique. The noble element (Cu) is kept in dilute concentration so that the rate of reduction of Cu is slow and limited by diffusion. The electrodeposition process is control- led by a computer which continuously integrates the charge during each layer deposition. The poten- tial is switched when the deposition charges for the nonmagnetic and the magnetic layers, QNM and QM, respectively, reach the set value. Such a procedure is required to give uniform layer thicknesses all along the filament. The electrodeposition process is stopped when the wires emerge from the surface (as evidenced by a sudden increase of the plating cur- rent). Dividing the membrane thickness (20 gm) by the number of cycles gives the average period for the wires involved in the GMR experiment. Using XEDS (X-rays Energy Dispersive Spectroscopy) microanalysis, the average chemical composition was determined. We can alternatively use another method based on the relationship between the layer thicknesses dNM and dM and the electric charges QNM and QM transferred at the cathode during the corresponding pulses. Fig. 1 shows the linear vari- ation of the bilayer thickness dbilayer as a function of

4 0 0 , i ~ i J

300L

200}-

f J

10

1 p I

E

0 02 04 06 08

Qcn (me)

Fig. 1. Relation between the mean thickness of the bilayer in Co/Cu multilayered nanowires dbilayer and the electric charge Oc, transferred at the cathode during Cu deposition at - 0.3 V. For this series, Oco is kept constant at 0.1 mC corresponding to a Co thickness of 24 nm. dbilayer is calculated using the ratio between the thickness of the membrane (20 gm) and the number of cycles required for the filling of the pores.

L. Piraux et al. / Journal of Magnetism and Magnetic Materials 175 (1997) 127 136 129

Qcu for the Co/Cu system. In this series, Qco is kept constant at a value of 0.1 mC and Qcu varies up to I mC; dbilayer was determined experimentally from electrodeposition process as explained above. Us- ing the Faraday's law, the following relationship can be derived:

dbilayer (nm) = 3.67 qcuQco + 34.3 qcoQco a n ' (1)

where A is the cathode surface area (in mm 2) which depends on the pore density and diameter; ~/cu and qco are the cathode current efficiencies for pulsed electrodeposition of Cu and Co, respectively; the two charges Qcu and Qco in Eq. (1) are expressed in inC. From the slope and intercept of the straight line of Fig. 1, we obtain qCo/r/cu = 0.72 and dco ~ 24 nm. Assuming a 100% current efficiency for Cu deposition, this leads to r/Co = 0.72. The microstructure of the single metal and multilayered

nanowires has also been studied in detail, using X-ray diffraction and analytical transmission elec- tron microscopy [6, 9, 12]. Fig. 2a shows a TEM image of a Co nanowire of diameter around 70 nm. The diffraction pattern is consistent with the HCP structure of cobalt with a c-axis oriented perpen- dicular to the wire axis. Fig. 2b shows a TEM image of a Nis0Fe20/Cu multilayer grown on a pure permalloy [6, 12]. It was shown that, in spite of the relative simplicity of growth by electrolysis, these multilayers exhibit single crystal structures, with single crystal grains including several tens of layers. The structure is FCC and the (1 1 0) axis is parallel to the growth direction. The layers look flat in the single crystal regions, while they can be quite distorded in the polycrystalline regions. The period of the superlattice shown in Fig. 2b is about 5 nm. The combination of TEM and EDX results tell us that the Cu layers are 1 nm thick and the permalloy ones are about 4 nm. It is noted that for a 20 gm

Fig. 2. (a) Transmission electron microscopy of a cobalt wire, in (1 0 0) projection. Dark field image and corresponding diffraction pattern. (b) TEM micrograph of a NisoFezo(4 nm)/Cu(l nm) multilayered nanowire grown on top of pure permalloy. The shadow on the left comes from another nanowire in the field of view. These micrographs are extracted from Ref. [6].

130 L. Piraux et al. / Journal o['Magnetism and Magnetic Materials 175 (1997) 127 136

thick membrane, multilayered nanowires com- posed of more than 3000 bilayers can be prepared using this method.

3. CPP-GMR in Ni80Fe20/Cu and Co/Cu multilayered nanowires

Magnetoresistance and magnetization curves obtained with magnetic fields parallel to the layers at 4.2 K on NisoFe2o(12 nm)/Cu(4 rim) multilayered nanowires are shown in Fig. 3a and Fig. 3b [-6]. The GMR ratio in the virgin state approaches 80% at low temperature which is at least a factor 20 larger than the values reported for Nis0Fezo/Cu multilayers in the CIP (current in the planes) geometry for comparable Cu layer thicknesses 1-13, 14]. The MR at the peak (73%) is somewhat smaller than in the virgin state. A similar relative

difference of less than 10% between the two mag- netoresistance ratios was also observed on Co/Cu multilayered nanowires I-4, 5]. At room temper- ature, the MR is reduced by a factor of 3. The saturation and coercive fields extracted from the magnetization curves are similar to those observed in GMR measurements. The large saturation fields observed for magnetic fields parallel to the layers are due to the demagnetizing fields arising from the multilayered nanowire structure.

In order to reduce the saturation fields, we have recently proposed another structure composed of Nis0Fe2o(3 nm)/Cu(10 nm)/NisoFe20(3 nm) tri- layers magnetically isolated by thick layers of Cu (100nm or more). A typical magnetoresistance curve for in-plane field is shown in Fig. 3c. For such structure, the expected magnetic behavior is that of isolated trilayers, essentially governed by their in-plane shape anisotropy. Our magnetization

80

70

t-- 60

"-" 50

N 4o ~ 3O

2O

10

0

I I i I I

-4 -2 0 2 4 H (kOe)

t J I I ] ]

(b) o . ,

-0.5

-6 -4 -2 0 2 4

H (kOe)

(e)

,,.., 15

10

5 -0

0 -4 -2 0 2 4 H (kOe)

(d)

0

I

'-6 -4

, f ,

J f l

-2 0 2 4 H (kOe)

< s z ~ l

_~_ i ~ : ~ NiF~

lOOnml !

! .fCu

Fig. 3. Giant magnetoresistance (GMR) and magnetization curves versus applied field parallel to the layers at 4.2 K for Ni8oFe2o(12 nm)/Cu(4nm) multilayered nanowires (a, b) and for multilayered nanowires composed of Ni8oFe20(3 nm)/Cu(10nm)/ NisoFe2o(3 rim) trilayers separated by 90 nm long Cu rods (c, d) (after Ref. [6]).

L. Piraux et al. /Journal of Magnetism and Magnetic Materials 175 (1997) 127 136 131

measurements have confirmed that the plane of the layers is an easy plane, with a saturation field of about 1 kOe (Fig. 3d). The dipolar interaction be- tween the two permalloy layers within a trilayer is also expected to favor an antiparallel orientation of their magnetic moment, which is in agreement with the almost zero remanent magnetization in Fig. 3d. The saturation field is definitely smaller than for the conventional multilayered nanowires of Fig. 3a and Fig. 3b. The large MR ratio (20%) that we observe is also consistent with a high degree of antiparallelism for the magnetizations of the two magnetic layers in a trilayer. The GMR curves are practically reversible and the maximum of resis- tance is reached at nearly zero field, which agrees with the very small value of the remanent magneti- zation and seems to confirm that the dipolar inter- action between the two NisoFe20 layers of a trilayer induces an antiparallel arrangement of their mag- netization at zero field. Also, there is almost no difference between the resistance at the peak of the MR curves and the virgin state, which also indi- cates that there is a unique zero field configuration induced by the dipolar forces.

We now present GMR measurements performed on electrodeposited Co/Cu multilayered nanowires and discuss the variation of the GMR with the thicknesses of the Cu and Co layers in the frame- work of the Valet-Fert (VF) model [7]. Two series of samples have been studied: Co(8 and 25 nm)/ Cu(10 nm ~< tcu ~< 350 nm), which we call series 1, and Co(60 n m < tco < 1 I, tm)/Cu(8 rim), which we call series 2.

,tN~ l~v) neglecting In the limit where ty, tF<</sf , 'sf, spin-mixing effects (low T regime) and using the general expressions of the GMR in the VF model [7], the MR is given by the following simple expres- sion [4, 5]:

AR']-I:2 _ p~tF + 2r* p~tN RAp) flp*tF + 27r* +/30*@ + 27r*"

This thickness dependence was first theoretically established in the simple case where R p and R aP represent the resistances of the parallel and antiparallel configurations. It has been later shown that the resistance R AP is the same for a strict antiparallel arrangement and when, less drastically,

AP refers to a state with a zero magnetization for the set of magnetic layers included in a thickness range of the order of the SDL [l 5]. Consequently, Eq. (1) holds also when the AP configuration is a random one. In a previous paper [5], we argue that the magnetizations in successive Co discs along a current line in samples of series 1 should be randomly oriented at low field (MR peak). We use the same notations as in Ref. [7], i.e. p~(,L) = 2p* in Cu, pT(+)= 2p*[1 - ( + ) / 3 ] in Co and p'~({)= 2r~[l - (+)7] at the interfaces where/3 and 7 are the bulk and interfacial spin asymmetry coeffi- cients, respectively.

Eq. (1) implies that in the limit tn, tF <<l~n)'Sf , 'sfttFI and at constant thickness of the magnetic layers, (AR/RAP) - 1/2 increases linearly with the thickness of the nonmagnetic layers. Such a linear variation is observed in our samples for t y (i.e. tcu) smaller than about 100-150 nm, as illustrated in Fig. 4 with data obtained at 77 K. Our results thus confirm the behavior systematically found by the MSU group in the long SDL limit [16, 17]. As predicted from Eq. (1), the slope of this linear variation decreases as tv (i.e. tco) increases. In addition, for ,, > fi, the straight lines corresponding to different values of tv are expected to pass through a single point hav- ing/3 1 as ordinate. On the contrary, for 7 < fl, the straight lines would not cross at a positive value of tN. The behavior observed in the inset to Fig. 4 is thus characteristic of 7 > fi and also allows us a di- rect determination of/3 (fi ~ 0.36).

As discussed in Refs. [4, 5], the determination of all the other parameters, * * "* DN, ,OF, t b and 7 involved in Eq. (1) was achieved using the slopes and intercepts of the two straight lines. We obtain: P*o ~ 1 8 x l0 -S f~m , /3-,~0.36, rco/cu* ~ 3 x 1 0 - 1 6 f 2 m 2, 7 ~ 0 . 8 5 a n d p * u ~ 3 X 1 0 Sf~m.

Out of the long SDL limit, deviations from the linear variation of Eq. (1) are expected by the VF model [7], with, for (NI tN>>/~f , an exponential de-

(N) crease of the GMR ratio as exp(--tN/lsf ). The ex- pected change in the plot of (AR/RAP) -1/2 vs. tN appears clearly in Fig. 4 for tcu ~> 150 nm. The dashed line in Fig. 4 has been calculated from the general expression of the GMR in the VF model [7] with the values of P'u, P~o, rco/cu,* /3, 7 derived from the long SDL regime (see above) and l(cu) ~ = 140 nm.

132 L. Piraux et al. /' Journal of Magnetism and Magnetic Materials 175 (1997) 127 136

,-. ~ 15

10

I

°o 350

6 o o / / O

/ / ©

0 40 80 120 160 / O / A_-J

5; 1;0 l ; 0 2;0 2;0 3;0

tcu (nm)

Fig. 4. Plot of (AR/R AP) 1,2 vs. tcu at T = 77 K for Co(8 nm)/Cu(tc,) multilayered nanowires. The dashed line is a fit of the results obtained by using the full expressions of the VF model (for more details, see Refs. [4, 5]). The inset shows also the data obtained on Co(25 nm)/Cu(tcu) multilayered nanowires (black circles).

e _

150

~, 100

50

T = 295K/F 7 /

/A ~/n • = 77K

JA

, , , L 0 200 400 600 800 1000

tCo (nm) Fig. 5. Linear variation of the inverse magnetoresistance vs. tco for Co(tco)/Cu(8 nm) multilayered nanowires at two different temperatures (after Ref. [5]).

Now, we present our results for series 2, with tcu = 8 nm and tco varying between 60 and 950nm (Fig. 5). For tv>>l{[ ) with tN<< IIN),,f and p*tN, -b "* \ \ *'F'~f~ ~*lW), the general expression of the GM R in the VF model, for example Eqs. (40~(42) in Ref. [7], becomes

AR /~2/(F) H ~ s f

R e - (1 - - fi2)t F' ( 2 )

where AR is the resistance change between random and parallel configuration (in the limit we are con- sidering, AR would be twice for a perfectly anti- parallel configuration [18]). In Fig. 5, we have plotted RP/AR vs. tco for series 2. The linear vari- ation we find agrees with Eq. (2). This agreement probably indicates that all the samples of series 2 present a similar random magnetic arrangement at the MR peak {more precisely, the condition for Eq. (2) is that the relative orientation of the mag- netizations at the two interfaces of Cu layers is randomly distributed [5]). At T = 77 K, by intro- ducing fl 0.36 in Eq. (2), we derive /(co~ = ~s |" =

44 + 4 nm. At 300 K, the slope is approximately twice that at 77 K. This could be due to some shortening of ~(Co) ,se at 300 K but also to some reduc- tion of the spin asymmetry coefficient fi as the temperature increases.

4. Anisotropic transport and magnetic properties of arrays of Co, Ni, Ni8oFe2o and Fe sub-micron wires

In this section we present data obtained on ar- rays of Ni, Co, Fe and NisoFe20 wires prepared using membrane samples of low porosity. For such arrays, the average spacing between the wires is

L. Piraux et al. / Journal of Magnetism and Magnetic Materials 175 (1997) 127 136 133

2500

2000

1500

~. lOOO

500

1000 1

0,8 - ( a ) 8oo

600 ~ 0,6 I . d

400 ~ 0,4

~ ~ 200 0,2 ¢__

• e- ~ - - i - l l 0 0 0 100 200 300 400 500

diameter (nm)

' ' ' i b )

I I I I

100 200 300 400 500

diameter (nm)

Fig. 6. Room temperature coercivity and squareness as a function of Ni (open squares) and Co (full squares) sub-micron wires diameter when the external field is applied parallel to the wires (after Ref. [9]).

much larger than 1 gm so that the dipolar interac- tions between wires are expected to be small [9, 19] and the magnetic properties are not significantly different from those of isolated wires. Fig. 6a and Fig. 6b show the variation of the coercive field and remanent ratio Mr/M~ for the Co and Ni-based systems as a function of wire diameter. The external field is applied along the axial direction. All wires exhibit a large increase of the coercivity as the diameter decreases. In contrast, a completely differ- ent variation of the remanent magnetization as a function of diameter is observed for Co and Ni wires. For the arrays of Co wires, the remanent ratio decreases with the diameter from a value of about 1 when q5 = 35 nm to 0.1 when ,;b = 500 nm. For infinite cylinder the shape anisotropy is 2~Ms (8200 Oe for Co) and tends to align the magneti- zation along the axis of the cylinders. When the magnetocrystalline anisotropy is small compared to the shape anisotropy, a square loop hysteresis is expected in the magnetization curve when the field is applied along the axis of the cylinder. We observe this behaviour only for Co nanowires with 4) = 35 nm [9]. The decrease of remanence for lar- ger diameter shown in Fig. 6b can be understood in the light of observations in the transmission elec- tron microscope (TEM) showing that the c-axis is oriented perpendicular to the wire axis (see Fig. 2a). As evidenced in previous works [9, 19], the contribution of the large magnetic anisotropy along the c-axis (of the order of K1 = 5 x 106 erg/cm a) perpendicular to the wire, competes directly with the shape anisotropy 0zM 2 = 6 x

106 erg/cm 3) which is along the axis of the wire. The effect of the crystal anisotropy is also evidenced by the decrease of the remanent magnetization. Fur- thermore, the decrease of the coercive field com- bined with the decrease of the remanent field when the diameter of the wires increases suggest that the wire tends to split into domains. Indeed, the forma- tion of a multi-domain structure depends on reduc- ing the magnetostatic energy at a cost of creating domain walls transverse to the axial direction. This is further confirmed by magnetoresistance mea- surements performed at low temperature.

Fig. 7a and Fig. 7b compare the magnetization hysteresis loops and the magnetoresistance for an array of Co wires with ~b = 90 nm. In our experi- ment, the current was along the axis of the wires. The magnetoresistance exhibits a positive vari- ation when the field is applied parallel to the cur- rent, while the effect is reversed when the field is perpendicular to the current. This can be under- stood in the framework of the anisotropic mag- netoresistance (AMR) phenomenon related to the change in the orientation between the magneti- zation and the current. These results also corrobo- rate the presence of domains within the wire with no preferential direction of the magnetization. The same experiments were also performed on arrays of Co sub-micron wires with ~b = 35, 200 and 450 nm [9].

In contrast, for arrays of Ni, Fe or Nis0Fe20 sub-micron wires, the magnetocrystalline aniso- tropy is no more comparable to the shape aniso- tropy (K1 is about one order of magnitude smaller

134 L. Piraux et al. /' Journal o f Magnetism and Magnetic' Materials 175 (1997) 127 136

than the shape anisotropy). As a consequence, the magnetic and AMR properties obtained on such systems are drastically different. As shown in Fig. 6b, the remanent magnetization of Ni wires

was always about 80% of the saturated magneti- zation for the whole range of wire diameters. The magnetization curves shown in Fig. 7c, Fig. 7e and Fig. 7g for arrays of Ni, NisoFe2o and Fe wires,

j (a) . , . . I ~ / /

/ I

/

-0,5 i

i / S -lO -5 0 5 lO

H (kOe)

1,Ol

1,005

0,995

0,99

(b)

-s 6 5 lO H (kOe)

1 I , (c) i

0,5!

(0L i

"0,51

- 1 ~ -10 -5

1,005, ' (d)

1

• 0,995~ Q...

0,99 // /

0'985!0 -iO

o H (kOe)

/ d

/ /

H (kOe)

xx

xx

10

, i

10 20

1

0.5

o

-0.5

(e)

/ / ( /

"110 -5

1.02

1

c:Z 0.98

0.96

0.94 -10 25

~ : = k.

S lO H(kOe)

t

(f)

/

/ / J i

/ /

/ / ",

6 ~o H(kO¢

1 f

o.5 (g) I

i'

-o . s / / / /

/

"115 -10 -5 0 5

H(kOe)

/ / / /

/ /

r I'

I 1~0 15

1,001 (h)

~. 0,999

/ .~" 0'i(7.

0,998 '.]i';" ':'i'i:'" :

t ""] 0'99710 -5 0 5 10

H(kOe)

Fig. 7. Hysteresis loop at room temperature and magnetoresistance at T = 4.2 K for an array of 90 nm Co (a, b), Ni (c, d), NisoFe20 (e, f) and Fe (g, h) sub-micron wires. The field is applied perpendicular (dashed line) and parallel (full line) to the wire axis.

L. Piraux et al. / Journal of Magnetism and Magnetic Materials' 175 (1997) 127 136 135

respectively, with 4) = 90 nm indicate that the easy axis is along the axial direction. This is further confirmed by the magnetoresistance measurements (see Fig. 7d, Fig. 7f and Fig. 7h) which have shown no variation of resistivity when an external field was applied parallel to the wire axis. So, it appears that in contrast to the cobalt case, information about the reversal process of the magnetization cannot be directly deduced from the resistivity measurements. Indeed, under formation of do- mains, no change of resistivity is expected for a par- allel field since the magnetization within each domain lies along the axial direction or parallel to the current. A detailed study of the magnetization reversal process in such arrays of sub-micron mag- netic wires was recently performed by means of magnetization, torque and magnetic force micro- scopy experiments together with micromagnetic calculations. The results are published elsewhere [10, 19].

situation is different in Co-based system because of a competing crystal anisotropy effect.

Acknowledgements

We thank many colleagues, especially J.M. Be- uken, R. Legras, E. Ferain, J.M. George, J.L. Maurice and R. Ferr6 for their contribution to parts of the work. We thank Whatman s.a. (Be- lgium) for providing some polycarbonate mem- brane samples used in this study. L.P. is a Research Associate of the National Fund for Scientific Re- search (Belgium). This work has been partly sup- ported by a Brite programme of the European Commission (BE95-1761) and by the Belgian In- teruniversity Attraction Pole Program (PAI-IUAP P4/10).

References

5. Conclusions

In summary, the template method and elec- trodeposition techniques were employed for generating arrays of nanowires of magnetic metals and multilayers in porous polymer membranes. Multilayered NisoFezo/Cu nanowires have shown promising GMR properties. In conventional multi- layered structure, GMR ratio as high as 80% were observed at fairly high field. Large GMR ratio at relatively low field was also obtained on a new structure composed of trilayers separated by thick Cu layers. An interesting control of the mag- netic moment arrangement by dipolar interaction has been evidenced. In the case of Co/Cu multi- layered nanowires, the interface and bulk spin- dependent scattering parameters as well as the spin diffusion lengths in the nonmagnetic and ferromag- netic layers are extracted from the GMR result analysis.

Finally, we have made a comparative study of the anisotropic magnetoresistance and magnetic properties in arrays of Co, Ni, Nis0Fe2o and Fe sub-micron wires. The shape anisotropy of indi- vidual wires dominates the magnetic properties in the Ni, Fe or NisoFe2o-based system while the

[1] W.P. Pratt, S.F. Lee, J.M. Slaughter, R. Loloee, P.A. Schroeder, J. Bass, Phys. Rev. Lett. 66 (1991) 3060.

[2] L. Piraux, J.M. George, J.F. Despres, C. Leroy, E. Ferain, R. gegras, K. Ounadjela, A. Fert, Appl. Phys. Lett. 65 (1994) 2484.

[3] B. Voegeli, A. Blondel, B. Doudin, J.Ph. Ansermet, J. Magn. Magn. Mater. 151 (1995) 388.

[4] L. Piraux, S. Dubois, C. Marchal, J.M. Beuken, L. gilipozzi, J.F. Despres, K. Ounadjela, A. Fert, J. Magn. Magn. Mater. 156 (1996) 317.

[5] L. Piraux, S. Dubois, A. Fert, J. Magn. Magn. Mater. 159 (1996) L287.

[6] S. Dubois, C. Marchal, J.M. Beuken, L. Piraux, J.L. Duvail, A. Fert, J.M. George, J.L. Maurice, Appl. Phys. Lett. 70 (1997) 396.

[7] T. Valet, A. Fert, Phys. Rev. B 48 (1993) 7099. [8] See, for example, T.M. Whitney, J.S. Jiang, P. Searson~

C. Chien, Science 261 (1993) 1316 and references therein.

[9] L. Piraux, S. Dubois, E. Ferain, R. Legras, K. Ounadjela, J.M. George, J.L. Maurice, A. Fert, J. Magn. Magn. Mater. 165 (1997) 352.

[10] K. Ounadjela, R. Ferr6, L. Louail, J.M. George, J.L. Maurice, L. Piraux, S. Dubois, J. Appl. Phys. 81 (1997).

[11] E. Ferain, R. Legras, Nucl. Instr. Meth. B 82 (1993) 539; Nucl. Instr. Meth. B 84 (1994) 331.

[12] J.g. Maurice, D. Imhoff, P. Etienne, O. Durand, S. Dubois, L. Piraux, J.M. George, P. Galtier, A. Fert, J. Magn. Magn. Mater., submitted.

[13] S.S.P. Parkin, Appl. Phys. Lett. 60 (1992) 512.

136 L. Piraux et al. / Journal o f Magnetism and Magnetic Materials' 175 (1997) 127-136

[14] S.K.J. Lenczowski, M.A.M. Gijs, R.J.M. Van De Veer- donk, J.B. Giesbers, W.J.M. De Jonge, Mater. Res. Soc. Symp. Proc. 34 (19951 341.

[15] S. Zhang, P.M. Levy, Phys. Rev. B 47 (1993) 6776. 1-16] P.A. Schroeder, J. Bass, P. Holody, S.F. Lee, R. Loloee,

W.P. Pratt, Q. Yang, Mater. Res. Soc. Syrup. Proc. vol. 313, Materials Research Society, 1993, p. 43.

[17] W.P. Pratt, S.F. Lee, P. Holody, Q. Yang, R. Loloee, J. Bass, P.A. Schroeder, J. Magn. Magn. Mater. 126 (1993) 406.

[18] L. Piraux, S. Dubois, A. Fert, in preparation. [191 R. Ferr6, K. Ounadjela, J.M. George, L. Piraux, S. Dubois

submitted for publication.