Exchange coupling and magnetoresistance in CoFe/NiCu/CoFe spin-valves near the Curie point of the spacer

0.25 MB·English

by  S. Andersson

Exchange coupling and magnetoresistance in CoFe/NiCu/CoFe spin-valves
near the Curie point of the spacer
S. Andersson and V. Korenivski
Nanostructure Physics, Royal Institute of Technology, SE-106 91 Stockholm,
Sweden
(Dated: 30 January 2010)
Thermalcontrolof exchangecoupling betweentwo stronglyferromagneticlayersthrougha weaklyferromag-
netic Ni-Cu spacer and the associated magnetoresistance is investigated. The spacer, having a Curie point
0 slightly above room temperature, can be cycled between its paramagnetic and ferromagnetic states by vary-
1
ing the temperature externally or using joule heating. It is shown that the giant magnetoresistance vanishes
0
due to a strong reduction of the mean free path in the spacer at above ∼30% Ni concentration — before the
2
onsetofferromagnetism. Finally,adeviceisproposedanddemonstratedwhichcombinesthermallycontrolled
n
exchange coupling and large magnetoresistance by separating the switching and the read out elements.
a
J

0
I. INTRODUCTION strate thermally controlled exchange coupling sam-
3
ples with structure Cu(90)/Ni Fe (8)/Co Fe (2)/
80 20 90 10
] Thermal control in spintronic devices and MRAM ap- NixCu1−x(t)/Co90Fe10(5)/Ta(10) (thickness in nanome-
ci plications has in recent years been of great interest due ters) were deposited. Three different thicknesses, t =
s to the associated increase of stability and decrease in 10,20,30 nm, of the weakly ferromagnetic Ni-Cu alloy
l- powerconsumption1–3. Recently,thermallyexcitedoscil- wereusedforstudyingtheinterlayerexchangeinthesys-
r lationsinnanocontacts,reachingfrequenciesoftheorder tem. VariationinxwasobtainedbycosputteringNiand
t
m ofGHz, havebeen predicted4. In this model a ferromag- CuontoSisubstratesthathadbeencutinto90x10mm
. netic (FM) film is separated from a small FM grain by strips. In this way a compositional gradient was created
at a point contact having a diameter of a few nanometers. along the Si strips ranging from x = 0.2 to x = 0.9. By
m Due tothe highcurrentdensitiesreachedinsuchasmall cutting the strips into smaller pieces a series of samples
- area,theFMregionwithinthepointcontactreachesvery with different Curie temperatures were obtained.
d hightemperatures. Whenthelocaltemperatureishigher Inordertoperformmagneticcharacterizationthesam-
n than the FM Curie point the exchange coupling through pleswereplacedinalooptracerequippedwithathin-film
o
the point contact becomes vanishingly small. heater. The temperature was controlled through a feed
c
[ Themodelisbasedontwopremises. First,athermally back loop using a type-T thermocouple in close contact
controlled exchange coupling between two FM regions. with the samples. Further investigations of the switch-
1 Second, an increase in resistance when the FM regions ing behavior at room temperature were performed using
v
decouple. Thefirstcriterioncanbemetinametallicsys- a vibrating sample magnetometer (VSM).
0
9 temwithtwostrongferromagnetsseparatedbyaweakly To measure the effect of Ni-Cu alloying on GMR,
0 FM spacer. If this can be combined with a large change samples with structure Ni80Fe20(4) / Co90Fe10(1) /
0 in resistance both criteria would be met. To date, the NixCu1−x(3.5) / Co90Fe10(2) were deposited. The Ni-
2. largest changes of resistance obtained in an all metal- Cu alloys were cosputtered from Ni and Cu targets such
0 lic structure have been from giant magnetoresistance5,6 that the stoichiometry, x, could be varied by controlling
0 (GMR). the relative difference in sputtering rates.
1 In this work we investigate the possibility of decou- Tomeasurecurrentinplane(CIP)GMR,thinAlwires
v: plingtwostrongferromagnetsseparatedbyaweaklyfer- were bonded to the top of the samples. Before electrical
i romagnetic Ni-Cu alloy. To verify if the Ni-Cu alloy, in measurements the separate switching of the NiFe/CoFe
X its paramagnetic phase, can be used as a GMR spacer bi-layer and CoFe top layer was confirmed using a mag-
r we study the effects of adding nickel to a copper spacer netometer.
a
in a spin-valve structure on the interlayer exchange cou-
pling and GMR. Finally, we design and implement an
improved thermionic spin-valve structure, in which the III. RESULTS AND DISCUSSION
switching and the read out layers are separated.
A Ni-Cu alloy was chosen for the weakly FM spacer
because of the well known dependence of Curie point
II. EXPERIMENTAL DETAILS on the Ni concentration7–9. The Ni-Cu spacer is used
to separate a magnetically softer NiFe/CoFe bi-layer
All films were deposited on thermally oxidized Si from a magnetically harder CoFe layer. Here NiFe and
substrates using magnetron sputtering at a base pres- CoFe stand for Ni Fe and Co Fe , respectively. By
80 20 90 10
sure better than 5 · 10−8 Torr. The argon pressure cosputteringNiandCu,aseriesofsampleswithdifferent
during sputtering was kept at 3 mTorr. To demon- Curie temperatures were obtained.
2
xx
1 (a) with that in Fig. 1 (b) in which the NiFe/CoFe
◦
1 (a) layeris heatedto 110 C andthereby decoupledfromthe
n CoFe layer, it can be seen that at roomtemperature the
o
it T = 25 oC } NiFe/CoFe layerhas notyetfinishedrotating180◦ when
a
z the hard layer switches. At point B the torque on the
it Irreversible
e CoFe layer is too strong for it to remain in position and
n
g an irreversible rotation of all layers follows. This behav-
a
M ior was confirmed by VSM measurements of the same
0
d B. sampleatroomtemperature. Startingatapositivefield,
ez Reversi}ble high enough so that all the moments were aligned, the
ila magnetizationwasmeasuredwhile the externalfieldwas
m
r swept to -20 Oe and then back again. After the field re-
xoN xversalat-20Oe the magnetizationbacktracksthe values
x-1 A. xmeasured before the reversal. The same behavior was
seen for field reversals up to -25 Oe indicating that the
-40 -20 0 20 40 rotation is reversible up to this point. For reversalfields
xx
Applied Field (Oe) any higher than this, the magnetization does not back-
xxtrack the values measured before the field reversal. This
1 (b) confirms that the switching behavior is irreversible for
n fields higher than ±25 Oe.
o
ita T = 110o C CoFe Fig. 1(b)showsthesamesampleat110◦C.TheCurie
z point of the spacer has been reached and the soft and
it
e hardFMlayersareessentiallyexchangedecoupledasev-
n
g idenced by the two distinct magnetization transitions at
a
M approximately 15 and 45 Oe. As the temperature is re-
0
d duced to room temperature the two magnetization tran-
e
z sitions shift towards each other and the sharp magneti-
ila zation loop becomes significantly skewed. At room tem-
m NiFe/CoFe
r peraturethecurveshapeisbacktotheoneshowninFig.
xo x
N 1(a). Thisthermallycontrolledinterlayerexchangecou-
x-1 xplingisperfectlyreversibleonthermalcyclingwithinthe
given temperature range.
-40 -20 0 20 40 ForsampleswiththeCuriepointaroundroomtemper-
Applied Field (Oe) ature the spacer had to be 20 or 30 nm thick in order to
completely diminsh the exchange coupling through the
FIG. 1. Normalized magnetization versus applied magnetic spacer. Anexplanationforthis couldbe thatthealloyis
fieldforasamplestructureNiFe/CoFe/NixCu1−x(30)/CoFe, not homogenousafter cosputtering at roomtemperature
x≈0.7at(a)roomtemperatureand(b)110◦C.Uponheating but contains regions with different Curie points. If the
theNi-Cualloygoesthroughaferromagnetictoparamagnetic spaceristoothin,theseregionscouldextendtothealloy
phase transition and the NiFe/CoFe bi-layer decouples from interfaces and couple the two CoFe films. Another pos-
theCoFe layer. sible explanation is that the alloying is homogenous but
the two strongferromagnetsarecoupled byexchangein-
teractionsthroughthespacerevenattemperaturesabove
A. Thermally controlled interlayer exchange coupling theCuriepoint. Ithasbeenindicatedthatexchangecan
propagatethroughparamagneticregionsonlengthscales
of several nanometers11, which is believed to be due to
Fig. 1 (a) shows the magnetization loop for a sam-
enhancement of magnetic order in thin layers caused by
ple with a 30 nm thick Ni-Cu layer having a Curie point
just above room temperature, x≈0.7. The shape of the the proximity effect of a strong ferromagnet.
curve indicates that the strongly FM layers are weakly
exchange coupled through the Ni-Cu alloy. Two distinct
regions can be seen — similar to the magnetic state of a B. CoFe/Ni-Cu/CoFe spin-valve
spring-magnet10. At point A in Fig. 1 (a) the magnetic
moment of the soft NiFe/CoFe layer starts to rotate in To understand the CIP GMR in the above
the external magnetic field. This is a reversible rotation NiFe/CoFe/Ni-Cu/CoFesystem,wehavetoconsiderthe
due to the exchange coupling through the Ni-Cu alloy mean free path in the Ni-Cu spacer. When the spacer
to the harder CoFe layer. The reversible switching con- thickness is much larger than the mean free path the
tinues until point B is reached. By comparing the mag- CIP GMR signal vanishes as exp(−t /λ)12. Here λ
NiCu
nitude of the magnetization at 25 Oe (point B) in Fig. isthe electronmeanfree pathinthe spacermaterialand
3
102 ther material optimization, we next examine how thin a
Calculated from data in [16] spacerwouldstillprovideameasurableGMRsignal. The
) Calculated from data in [17] thinnest possible spacer to avoid any significant RKKY
m Calculated from data in [18] coupling is ∼3 nm13. We therefore choose this thickness
n
( Our measurements and evaluate the CIP GMR versus Ni concentration.
h
ta101 We have fabricated samples with structure
P
e NiFe(4)/CoFe(1)/Ni-Cu(3.5)/CoFe(2). CIP GMR
e measurements for samples with different Ni concentra-
r
F
tions are shown in Fig. 3 (a). With pure Cu in the
n
a spacerasignalof3.7%ismeasuredatroomtemperature
e
M100 (5% at77 K). When Ni is addedto the spacer the signal
drops caused by a decrease in λ. At 29 at. % Ni the
GMR has decreased by more than a factor of three,
0 20 40 60 80 100
which we attribute to a sharp decrease of the mean free
Ni Concentration (at. %)
path on alloying. Fig. 3 (b) shows the magnetization
versus applied field for this sample. It can be seen that
FIG.2. ElectronmeanfreepathinbulkNi-CufordifferentNi the switching is still separate at this concentration.
concentrations. The data points have been calculated using
At 35 at. % Ni the magnetic layers couple and the
theDrudemodelfrom publisheddataonresistivitymeasure-
mentsat 300, 273 and 250K16–18. switching is no longer separate,which is shownin Fig. 3
(c). It can thus be concluded that even for a very thin
Cu-Ni spacer the CIP GMR signal essentially vanishes
6
for 29 at. % Ni concentration, where the layers are still
(a)
1(b)
decoupledmagnetically. The finaldecreasetozeroGMR
5 0 at35at. %Niis due tocoupling throughthe spacerand
) Ms-1 not to a decrease in λ.
%( R 4 77 K / M 1(c) peIrtpwenildlibcuelianrfotromthateivpelatnoepGoiMntRouintsttheaatdboyfuCsIinPgtchuerrliemnt-
M 3 0 iting length scale would be the spin diffusion length14,
G l , and not λ. However, published experimental data
P -1 insfdicate that l decrease with the same rate as λ and
IC 2 -100 0 100 has been measusfred to be 7.5 nm at 5◦K in an alloy with
300 K H (Oe)
22.7% Ni15. This is still below the minimum thickness
1
inourcase of20nmneededto achievereliableinterlayer
decoupling.
0
0 10 20 30 40 50 60 To overcome the above limitations and use thermally
Ni Concentration (at. %) controlled interlayer exchange coupling together with
large magnetoresistance we propose a new design in
whichtheGMRreadoutlayerandtheweaklyFMspacer
FIG. 3. (a) Current in plane (CIP) giant magnetoresistance
(GMR) versus Ni concentration in a NiFe(4)/Co2Fe(1)/Ni- are separated. A schematic is shown in the inset to Fig.
Cu(3.5)/CoFe(2) spin-valve. The inset shows the n2ormalized 4. An antiferromagnet (AFM) is used to exchange bias
magnetizationversusappliedfieldforaspin-valvew2ith(b)29 a FM film which works as a reference layer in the spin-
at.% Ni in the spacer and (c) 35 at. % Ni in the spacer. valve. The spin-valve uses a metallic spacer for GMR or
an insulator for tunneling magnetoresistance. The spin-
valve spacer is only used for read out and can be opti-
t is the spacer thickness. In order to obtain a high mizedtogivethe highestpossiblemagnetoresistancesig-
NiCu
GMR signal the thickness of the spacer should be com- nal. The top layer of the spin-valve is exchange coupled
parableorsmallerthanλ. Fig. 2showsλfordifferentNi througha weakly FM alloy to a top pinned FM. At high
concentrations from our measurements as well as calcu- temperatures the coupling through the weakly FM alloy
lated from published experimental data using the Drude is negligible and the top layer of the spin-valve is free to
model. The mean free path decreases quickly with in- rotate. However,atlowtemperaturesitwillbeexchange
creasing Ni concentration. In the interesting range of coupled to the top pinned FM. This results in full flexi-
Ni concentrations between 40% and 70% where the al- bility when choosing the composition and Curie point of
loy is weakly ferromagnetic7, λ is down to ∼1 nm. As the weakly ferromagnetic alloy while at the same time
wasdetailedintheprevioussection,ourspacerswiththe making it possible to achieve high magnetoresistance.
Curie point close to room temperature have a minimum To demonstrate this new struc-
thickness of 20 nm in order to completely diminish the ture we have deposited samples of
interlayer exchange coupling. Assuming the thickness of NiFe(3)/MnIr(15)/CoFe(4)/ Cu(3.5)/CoFe(4)/NiFe(6)/
the weakly ferromagnetic spacer can be reduced by fur- CoFe(2)/NiCu(20)/CoFe(2)/NiFe(3)/MnIr(12)/Ta(5)
xx
xx
4
spacerisdemonstrated. Attemperatureshigherthanthe
2.5 1.3·106A/cm2
1.0·106A/cm2 CuriepointofthespacertheFMfilmsaredecoupled. At
7.5·105A/cm2 lower temperatures the switching behavior can be sepa-
) 2.0 rated into two regions — reversible and irreversible.
%
xx
( R AFM In a CoFe/Ni-Cu/CoFe spin-valve the CIP GMR sig-
xM1.5 xnal vanishes due to a sharp reduction of the mean free
G path on alloying for Ni concentrations above ∼30 at. %.
Weak FM Alloy
P A new design is proposed and demonstrated, combining
xI1.0 x
C Metal / Insulator thermallycontrolledexchangecouplingandlargemagne-
xxtoresistance, which may prove useful for applications in
x0.5 xcurrent controlled magneto-resistive oscillators.
AFM
ACKNOWLEDGMENTS
0
-80 -60 -40 -20 0 20 40 60
This work was supported by EU-FP7-FET-STELE.
Applied Field (Oe)
FIG. 4. CIP GMR for three different current den-
sities versus applied field measured on a sample 1I. L.Prejbeanu, M.Kerekes, R. C.Sousa, H.Sibuet, O. Redon,
structure NiFe(3)/MnIr(15)/CoFe(4)/Cu(3.5)/CoFe(4)/ B.DienyandJ.P.Nozi´eres,J.Phys.Condens.Matter19,165218
NiFe(6)/CoFe(2)/NiCu(20)/CoFe(2)/NiFe(3)/ MnIr(12)/ (2007).
Ta(5). The inset shows a schematic of a device in which 2R.S.Beech,J.A.Andersson,A.V.PohmandJ.M.Daughton,
J.Appl.Phys.87,6403(2000).
thermally controlled exchange coupling is separated from 3J.M.DaughtonandA.V.Pohm,J.Appl.Phys.93,7304(2003).
a spin-valve read out. Either giant magnetoresistance or 4A.Kadigrobov,S.I.Kulinich,R.I.Shekhter,M.JonsonandV.
tunnelling magnetoresistance can be used for read out. Korenivski,Phys.Rev.B74,195307(2006).
5G.Binasch,P.Gru¨nberg,F.SaurenbachandW.Zinn,Phys.Rev.
B39,4828(1989).
with 70 at. % Ni in the spacer (T ≈ 100◦C). The 6M.N.Baibich,J.M.Broto,A.Fert,F.NguyenVanDauandF.
C
samples were patterned using photolithography into Petroff,Phys.Rev.Lett.61,2472(1988).
7T.J.Hicks,B.Rainford,J.S.KouvelandG.G.Low,Phys.Rev.
strips 50 µm wide and 1 mm long and then bonded Lett.22,531(1969).
at the edges with aluminum wire. The resulting CIP 8S. K. Dutta Roy and A. V. Subrahmanyam, Phys. Rev. Lett.
GMR signal for different current densities is shown in 177,1133(1969).
FIG. 4. When the current density is increased and the 9J. B. Sousa, M. R. Chaves, M. F. Pinheiro and R. S. Pinto, J.
Low.Temp.Phys.18,125(1975).
temperature of the device is correspondingly raised due
10J.E.Davies,O.Hellwig,E.E.Fullerton,J.S.Jiang,S.D.Bader,
tojouleheating,thetoplayerofthespin-valvedecouples G.T.Zima´nyiandK.Liu,Appl.Phys.Lett.86,262503(2005).
from the top pinned CoFe/NiFe bi-layer producing a 11A.HernandoandT.Kulik,Phys.Rev.B49,7064(1994).
strong GMR signal. For a current density of 1.3 · 106 12A.Barth´el´emyandA.Fert,Phys.Rev.B43,13124(1991).
A/cm2, T > T ≈ 100◦C, the exchange decoupling is 13S.S.P.Parkin,R.BhadraandK.P.Roche,Phys.Rev.Lett.66,
C
2152(1991).
complete and the full CIP GMR signal of 2.5% for this 14J.BassandW.P.PrattJr.,J.Phys.Condens.Matter19,183201
structure is obtained.
(2007).
15S. Y. Hsu, P. Holody, R. Loloee, J. M. Rittner, W. P. Pratt Jr.
andP.A.Schroeder,Phys.Rev.B54,9027(1996).
IV. CONCLUSION 16R. W. Houghton, M. P. Sarachik and J. S. Kouvel, Phys. Rev.
Lett.25,238(1970).
17H.M.AhmadandD.Greig,J.Phys.(Paris).35,C4-223(1974).
Thermally controlled exchange coupling between two 18P. A. Schroeder, R. Wolf and J. A. Woollam, Phys. Rev. 138,
strongly FM films separated by a weakly FM Ni-Cu A105(1965).