IEEE TRANSACTIONS ON MAGNETICS, VOL. 35, NO. 5, SEPTEMBER 1999
2952
Ion Beam Deposition and Oxidation of Spin-Dependent Tunnel Junctions Susana Cardoso '.,VCronique Gehannob, Ricardo Ferreira' and Paulo P. Freitas a INESC, R.Alves Redol, 9-1", 1000 Lisboa, Portugal
Abstract - Spin dependent tunnel junctions showing tunnel magnetoresistance (TMR) values of 39-41 % were fabricated using Ion Beam Deposition (IBD). Both the electrodes and the aluminum layer deposition were done by IBD. The aluminum oxidation was performed using the assist gun with an oxygen beam (+30 V acceleration voltage applied on the grids) using mixed 02/Ar plasma. The oxidation was monitored in real time with a residual gas analyzer (RGA). The junction area is defined by lithography, down to 3x2 pn2.As-deposited junctions with 15A of A1 showed TMR of 27-29 %, independent of the junction area, with resistance-area products of 0.8-1.6 MRxpn'. This TMR value reached 40 % upon annealing at 290"C, with resistance decreasing to 0.5-0.8 MRxpn2.
Index Terms -Ion beam depositiodoxidation, spin polarized tunnel junctions
I. INTRODUCTION
'
290°C increases the TMR signal up to 40 %, with a slight decrease in junction resistance.
11. EXPERIMENTAL METHOD AND RESULTS The electrodes and A1 layer deposition, as well as the oxidation, were made in an automated load-locked IBD system (Nordiko N3000) equipped with a 10 cm-diameter deposition gun and a 25 cm-diameter assist gun [6]. The IBD chamber has a base pressure of 5x10" Torr. A quadrupoletype residual gas analyzer (RGA) was connected to a sampling chamber for diagnostic of oxygen and argon levels inside the main chamber during process. Sampling of the chamber atmosphere along the process is made through a needle valve, that assures RGA operating at - l ~ l O - Torr ~ during all the steps of the junction deposition. Fig.1 shows a schematic representation of the depositiodoxidation chamber.
Recently, high room temperature tunneling magnetoresistance values (TMR 20 to 27 %) were reported by our group in micron-sized junctions with low resistance area product (1 to 10 kQxpm2), for junction areas ranging from 6 to 45 pm2 (tAl=7-13 A, oxidized 4 to 90s) [I]. TMR values up to 36% were obtained after junction anneal [2]. In our previous work, both bottom and top electrodes were prepared by magnetron sputtering, and the A1203 barrier was prepared by, depositing a thin A1 film (DC magnetron) followed by 0 2 plasma oxidation (RF) using two different machines, thus two vacuum breaks. These large TMR values have also been reported by few other groups, where junctions are prepared without vacuum break [3,4]. In this work, the Ion Beam Deposition (IBD) technique [5,6] is evaluated as a means for fabrication of spin tunnel junctions. The IBD method can produce higher quality thin films (less defects), making IBD a good candidate for the deposition of thin A1 films in the junctions. The IBD is also a Fig.1. Schematic drawing of the IBD system. Pictorial representation of the plasma during oxidation is shown. N1 and N2 are the neutralizers for versatile tool for the oxidation process, allowing good control deposition and assist guns, respectively. Sample distance to assist gun grid of ion kinetic energy. The oxygen level was scanned during and target are 30 cm and 20 cm, respectively. the deposition and oxidation steps with a residual gas analyzer (RGA) installed for this purpose. Results from micron-sized The fabricated junction structure is (see Fig.2): glass/ Ta junctions based on C O ~ ~ F ~ ~ ~ / A ~ ~(t~1=15 O ~ / A) C Oare X~F I ~ Ni~oFe20(70 A)/ Cos2FeI8(30 A)/ A1203/ CoszFels (90~ A)/ presented. TMR is 27-29 % at room temperature, (40 A)/ Mn7JrZ6 (250 A)/ Ta (30 A). The structure is independent of the junction area, with resistance-area protected with 150 A of sputter-deposited TiloW90(Nz). This products of 0.8-1.6 MQxpm2. Annealing of these junctions at layer acts as an antidiffusion barrier between the Manuscript received March 5 1999. S.Cardoso, 351- 1-3100380, fax 351-1-3145843;
[email protected] This work was supported in part by the projects PRAXIS P/CEX/FIS/28/96 and P/CTM/10220/1998. S.Cardoso was supported by the PRAXIS XXVBD/l1533/97 grant. " Also at Instituto Superior TCcnico, Av.Rovisco Pais, 1096 Lisboa, Portugal. Present adress: CENLETI, 17 Rue des Martyrs, 38054 Grenoble, France. 0018-9464/99$10.00 0 1999 IEEE
2953
'
AIl%Si0.4%Cu contact and the top Ta layer. The Mn-Ir exchange material was chosen because of its high blocking temperature ( T ~ = 2 5 0"C)[ 6 ] .For Ta, Ni-Fe, Mn-Ir and CO-Fe deposition, an acceleration voltage of +I450 V (29 mA Xe beam, at 3 ~ 1 0 Torr, ' ~ 1 sccm) and extraction voltage of -300 V were used. An aligning field of 40 Oe created by a permanent magnet array is mounted around the substrate table. The substrate table rotates at 15 rpm and is water cooled. The properties of 10008, thick films deposited in the conditions used for the junction layers are reported in Table.1. TABLE I PROPERTIES OF. IO00 A THICK FILMS
DEPOSITEDBY IBD deDosition rate
resistivitva
0.34 0.29 0.33 0.25
21.7
Ni-Fe CO-Fe Mn-Ir A1
Fig. 2. Junction structure.
17.1 175 4
a Measured using an in-line four probe dc method.
The A1203barrier was made by oxidation of the deposited AI layer (158,) for 60 s using a mixed 0 2 / A r beam. The beam was created by applying a voltage of +30 V to the assist grids, accelerating the ions from the plasma (RF plasma, 80 W). The Ar and O2gas flows are controlled separately by two different mass flow controllers. The plasma potential is around +20 V with respect to ground. In the absence of an acceleration voltage on the grids, this leads to a kinetic energy for the 0 and Ar ions of the order of 20 eV. Fig.3 shows the oxygen and argon profiles obtained in real time with the RGA, during the oxidation process. The oxidation follows three steps: a) starting the assist gun plasma at 100 W with Ar (8 sccm), b) injection of 40 sccm of oxygen, keeping 8 sccm of Ar, and c) the Ar flow is reduced to its final value (4 sccm), and oxygen remains at 40 sccm, with RF power decreased to 80 W. Voltages are applied to the grids, shutter is open and oxidation time starts counting. The pressure in chamber is 6 ~ 1 0 Torr - ~ during oxidation. The starting, step is needed since the pure O2 plasma is unstable and harder to start.
The plasma is also more stable if a small amount of Ar is left during oxidation. The ions observed with the RGA during oxidation are O+, O;, Ar'+ and Ar' (mass-electric charge ratio M/z=16, 32, 20 and 40, respectively). The partial pressures of the existing ions during the oxidation are: p(O;)=3.8x 10-' Torr, p( 0+)=6.8x10-9, p(Ar+)=2.3x1 0-' and p(Ar'+)=3.8~10-~, measured with the RGA at a total base pressure of Torr in the sampling chamber. The water content was not changed (p(Hz0+)=5.2~1O-~Torr) before and after oxygen is injected in the assist gun. Junction oxidation can be done with (+20 to +50 V) or without (0 V) acceleration voltage on the positive grid. No differences are observed in the sampled ions relative amount in case a small acceleration voltage was applied, instead of a RF plasma only. At the end of oxidation (the assist gun turns off) the amount of A d 0 ions in the chamber reduces sharply and immediately, and the processing of the top electrode continues after cleaning oxidized target surfaces (-2 minutes pre-sputtering of each target). In this oxidation process the plasma is created away from the substrate, contrary to the previously reported INESC process [ 1, 21, where the sample is immersed in the plasma. Fig. 4 summarizes the deposition and oxidation process, for a batch of 8 wafers. Bottom electrode and A1 are deposited for the 8 samples, then the oxidation is done for the batch, and samples moved into the load-lock ( 2 ~ 1 0 .Torr), ~ while top electrode targets are pre-sputtered. The junction fabrication finishes with top electrode batch deposition. The junctions were patterned using a self-aligned process by direct laser lithography, ion milling and lift-off, defining junction areas down to 3x2 pm2 [I, 21. A via in sputterdeposited Si02 defines the junction area, and a 3000 8,thick 4
32
8x 28
g
h
24
I
3 K
'E ",
20
1E-5
I
k IE-6
I
top electrode targets cleaning
We)
I
samples in loadlock (p=2xlO~'Torr)
v
-
latency tim
(-30s)
36
16 L
12
0
2
Q
1E-7
8
3
1E-8
2 n
;
4
IE-9
-- -
lE-l0l 1E-11 i
0
-!
10
-
0
shutter open
,
i(6.09).
,
,
,
,
,
,
,
. ..... , , , , 1
20 30 40 50 60 70 80 90 1 0 0 1 1 0 Time [SI
Fig.3. Time evolution of the O', Oz', Ar" and Ar' ions partial pressures during oxidation, obtained by the RGA in the sampling chamber.
1 E-5
1 E-4
1 E-3
C h a m b e r pressure (Torr)
Fig.4. Summary of the deposition and oxidation processes, for a sequence of 8 wafers.
2954
AIl%Si0.4%Cu film is used as top contact. The bottom free electrode dimension is 200x17 pm’. Junction thermal anneal was made in vacuum (p-1x10-6 Torr), during 45 minutes at a temperature T A , with a rising time of 90 min. A magnetic field of 500 Oe,was applied along the easy direction of magnetization during anneal and furnace cooldown. Fig. 5 shows TMR results for junctions with areas of 2x12 pm’. Free layer coercivity is 8 Oe. For all fabricated junctions with dimensions from 2x3 pm2 to 70x70 pm2,the TMR value for the as-deposited junction is 27-29 %. Independent of the junction size, the average junction resistance-area product RxA is 0.8- 1.6 MClxprn’. After annealing at 290°C the TMR increases to 39-41 %, while RxA decreases to 0.5-0.8 MRxpm’. The inset of Fig.5 shows the I-V curves for the asdeposited junction (2x12 pm’), and after anneal at 290°C. The positive and negative branches of the I-V curves are similar, both for the as-deposited and annealed sample, which indicates no partial oxidation of the bottom electrode [l]. From the fitting of the curves, barrier height increased from 2.15 eV in the as-deposited junction to 2.43 eV after annealing at 290”, while the effective thickness decreased from 10.3 A to 9.0 A.
40 -
....
10
0
20
300 600 900 12001500 Voltage [mV]
Annealed @ 2
Junction area:
As-deposited
k 4 6 . 7 kR,TMR=40 %
-100
;l , o C
m
; .-;0,5 U)
a,
I
U
25
‘ i s 0 200 240 280 320 Annealing temperature [QC]
-50
Ion Beam deposition and oxidation of tunnel junctions with high TMR and high thermal stability was successfully demonstrated. Both the electrode deposition and the oxidation process were performed in the same working chamber, without vacuum break. The AI oxidation was made u s i n g an Oz/Ar b e a m , w i t h a v o l t a g e of t30 V a p p l i e d t o t h e
). R=22.8 kQ,TMR=28.2%
-0-
1,5
a
111. CONCLUSIONS
..I..-.-.-.-.-..
30 . 0-0-0.0-0-0-0-0-0-025 15
4I Fig.6. Dependence of TMR and resistance-area product on the annealing temperature for IBD junctions.
35
’ i-.-
junctions the barrier is robust and that TMR values of 4-6 % still remain after anneal at 400°C [81.
0
50 Applied Field [Oe] ,
100
150
Fig.5. TMR curves for 2x12 ‘m2 IBD junctions, as deposited, and after annealing at 290°C. The inset shows the I-V curves, from where the . 1 5(9.0 8, effective thickness ~f,=10.38, and the barrier height ~ D ~ f ~ 2eV and 2.43 eV after annealing) were obtained using the Simmons model.
The TMR increase upon high temperature anneal has been previously reported by us [7] on junctions prepared by PVD and plasma oxidized. For junctions with resistance-area products of 10-13 MRxpm’, TMR increased from 22 to 26% for annealing temperatures close to 200”C, with small changes in the barrier (aeff decreases 0.2 eV). For junctions with relatively low as-deposited resistance-area products (25-35 kRxFm2), barrier height ,increased and effective thickness decreased with anneal [7]. The present junctions prepared by IBD show better thermal stability, withstanding thermal treatment up to 3 1O”C, with continuously increasing TMR. Fig.6. shows annealing temperature dependence of the TMR signal and junction resistance, for the different junction areas. Further anneals at higher temperatures show that for IBD
assist grid. The as-deposited junctions show TMR values of 27-29 %, and after anneal at 290°C this value increases to 39-41 %. REFERENCES [l] J.J.Sun, P.P.Freitas and V.Soares, “Low resistance ,spin-dependent tunnel junctions deposited with a vacuum break and RF plasma oxidized’, Appl. Phys. Lett, vo1.74, pp.448-450, January 1999. [2] R.C.Sousa, J.J.Sun, VSoares, P.P.Freitas, A.Kling, M.F.da Silva, J.C.Soares, “Large ,Tunneling Magnetoresistance Enhancement by Thermal Anneal”, Appl.Phys.L.ett, ~01.73, pp.3288-3290, November 1998. [31 SXParkin, “Spin dependent tunneling and its application to nonvolatile magnetic random access memory”, J.Appl. Phys., vo1.85, pp. 5828-5833, April 1999. [41 S.Tehrani, E.Chen, M.Durlam, J.M.Slaughter, and J.Shi, “High density submicron magnetoresistive random access memory”, J. Appl. Phys., ~01.85,pp.5822-5827, April 1999. [5] M.Tan, “Ion Beam deposition: meeting the challenge of.thinner films”, ’ Data Storage, pp.35-38, January 1996. [6] V.Gehanno, P.P.Freitas, A.Veloso, J.L.Ferreira, B.Almeida, J.B.Sousa, A.Kling, J.C.Soares and M.F.da Silva, “Ion Beam deposition of Mn-Ir Spin Valves”, accepted to publication, E E E Trans.Magn. (May 1999). [7] R.C.Sousa, J.J.Sun, V.Soares, P.P.Freitas, A.Kling, M.F.daSilva, J.C.Soares, “Temperature Dependence and Annealing Effects on Spin Dependent Tunnel Junctions”, J. Appl. Phys., ~01.85,pp.5258-5260, April 1999.. [81 S.Cardoso and P.P.Freitas, to be published.
