Magnetoresistance of spin-dependent tunnel junctions with composite electrodes

(1)

Magnetoresistance of spin-dependent tunnel junctions with composite electrodes

C. H. Ho and Minn-Tsong Lin^a)

Department of Physics, National Taiwan University, 106 Taipei, Taiwan Y. D. Yao and S. F. Lee

Institute of Physics, Academia Sinica, 115 Taipei, Taiwan C. C. Liao, F. R. Chen, and J. J. Kai

Department of Engineering and System Science, National Tsing-Hua University, 300 Hsinchu, Taiwan 共Received 2 March 2001; accepted for publication 17 September 2001兲

Spin-dependent tunnel junctions, Co/Al₂O₃/Co 共CoFe兲/NiFe, were fabricated to investigate the effect of the additional Co 共CoFe兲 interlayer on tunneling magnetoresistance. The quality of the junction was examined with a cross-sectional image generated by high-resolution transmission electron microscopy, and an electron energy loss spectra map. For junctions with a Co 共CoFe兲 interlayer in the top electrode thinner than 0.8 nm 共1.0 nm兲, the tunneling magnetoresistance ratio increases with interlayer thickness. For junctions with a 0.8 –2.0 nm Co 共1.0–2.0 nm CoFe兲 interlayer in the top electrode, the tunneling magnetoresistance ratio reaches the maximum value of 2.16 共4.45兲 times that without any Co 共CoFe兲 interlayer in the top electrode. The increase in the tunneling magnetoresistance ratio may be attributed to the increased effective ferromagnetic electrode polarization and the various spin-flip scattering factors. © 2002 American Institute of Physics. 关DOI: 10.1063/1.1419259兴

I. INTRODUCTION

Spin-dependent tunnel共SDT兲 behavior between a pair of ferromagnetic layers separated by an insulator layer are cur- rently the subject of much research.¹ These behaviors are fundamentally important in understanding spin-polarized electron tunneling and applications in digital devices, such as sensors and magnetic random access memory共MRAM兲. The tunneling magnetoresistance 共TMR兲 ratio is defined as (R_↑↓

⫺R_↑↑)/R_↑↑, where R_↑↑ and R_↑↓ are the resistances for par- allel and antiparallel spin alignment states of two ferromagnetic layers in the SDT junction, respectively. The ratio can be generally described by Julliere’s model:²

TMR ratio⫽ 2 P₁P₂

1⫺P1P₂⫻100%. 共1兲

Where, P₁ and P₂ are spin polarization values of con- duction electrons in two ferromagnetic electrodes共FM elec- trodes兲. Many details affecting TMR, such as the interfacial effect, are not yet been fully understood due to difficulties in fabricating and controlling the quality of SDT junctions. The spin-flip scattering between tunnel electrons and impurities in the insulator or at the insulator–ferromagnet interface rep- resent an important effect. The presence of impurities, such as Co-, Ni-, Cu-, and Pd-based ions, in the insulator layer of Co/Al₂O₃/Ni₈₀Fe₂₀SDT junctions has been reported to cause a reduction of the TMR ratio, due to the spin scattering.³ Another important interfacial effect is the increased polarization of FM electrodes by adding high-polarization materials

at the insulator–ferromagnet interface. This increase follows from the influence of the ferromagnet-insulator coupling on the effective polarization of the FM electrode, as indicted in an earlier theoretical work.⁴A higher polarization value of the additional interlayer yields a higher effective polarization value of the FM electrode, and therefore a higher TMR ratio.

Jansen et al. have reported an increase in the TMR ratio by up to 1.25 times by adding Fe-based ions in the insulator of SDT junctions.⁵

This article presents the influence on TMR, as a function of thickness, of different additional interlayers, Co and CoFe, deposited at the interface between the insulator and the FM electrode of SDT junctions. The TMR ratio was dramatically increased by both Co and CoFe interlayers in the low interlayer thickness range, but was reduced at higher coverages of FM interlayer.

II. EXPERIMENTAL DETAILS

All SDT junctions were prepared in a high-vacuum mag- netron sputtering system with a base pressure below 3

⫻10^⫺7Torr. Three contact masks were employed to deposit the rectangular bottom Co electrode layer 共10 nm兲, the cir- cular insulator Al₂O₃ 共2.3 nm兲, and the rectangular NiFe electrode layer共10 nm兲 on 7059 Corning glass substrates. Co and NiFe were deposited by dc power with a deposition voltage of around 300 V in 5 mTorr of Ar ambient. The insulator layer was formed using RF glow discharge 共64% Ar ⫹ 36%

O₂) with a ⫺350 V bias voltage on the Al film. The additional interlayers Co and CoFe were grown at a slow deposition rate 共0.02–0.05 nm/s兲 on the Al2O₃ layer. A 50:50 CoFe alloy was used here.

a兲Author to whom correspondence should be addressed; electronic mail:

[email protected]

JOURNAL OF APPLIED PHYSICS VOLUME 90, NUMBER 12 15 DECEMBER 2001

6222

Downloaded 23 Dec 2001 to 140.112.101.91. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp

(2)

A cross strip junction with 1 mm⫻1 mm area was fabricated for the four-probes measurement of the tunnel resistance in a current perpendicular to the film plane 共CPP兲.

Electrical properties were measured using a dc source at room temperature. Magnetic properties, such as the switching field, were determined by a longitudinal magneto-optical Kerr effect 共MOKE兲 instrument.

The electron microscopic observation of the cross section of the SDT junction first followed standard thinning procedures.⁶ The sample was mechanically polished, dimpled to a thickness of 3␮m, and then ion milled共usually at 5 keV, 1 mA兲 to perforation. The image of the cross- section image was taken by a JEOL 2010F field emission gun high-resolution electron microscope共HRTEM兲 equipped with an Oxford energy dispersive x-ray spectrometer共EDS兲.

All images were obtained at an electron accelerating voltage of 200 kV.

Electron energy loss spectroscopy共EELS兲, in a reflective geometry, is a powerful tool with which for probing electron excitation at surfaces of ultra-thin films, with an element sensitivity which yields nanometer spatial resolution. The in- cident electron beam was normal to the film plane in TEM and EELS measurements. EDS and EELS were conducted with a 0.5 nm nanobeam probe.

III. RESULTS AND DISCUSSION

The tunnel resistivity, that is the RA value defined as the product of the tunnel resistance R and the junction area A, of the SDT junctions, glass substrate//Co 共10 nm兲/ Al2O₃ 共2.3 nm兲/Co and CoFe/NiFe 共10 nm兲, varied from 10 to 100 M⍀␮^m². The tunneling properties of SDT junctions were examined by I – V measurement. Figure 1 plots the I – V curve of the SDT junction with the additional Co interlayer of 2.0 nm. This curve can be fitted by Simmons’ formula,⁷ J⫽␣^V⫹␥^V³, to give the effective barrier width, d, and height, ␾, where J is the current density through the SDT junction and V is the bias voltage across two FM electrodes.

Two coefficients, ␣ ^and ␥, are functions of d and ␾^{. The} fitting results show that all of the SDT junctions considered here have a ␾ of 1–2 eV and a d of 2.3–2.5 nm. These values are consistent with those of earlier research,^8,9imply- ing sufficient insulator quality.

Figure 2 depicts the cross-sectional TEM image of the SDT junction with the additional 1.2 nm thick CoFe interlayer. Both the top and bottom FM electrodes are polycrys-

talline, as commonly observed in sputtered thin films. The interfacial root-mean-square roughness of the individual layer is relatively low, only 0.22–0.34 nm, suggesting that the pinhole may not exist, and that no shunting current, other than the tunneling current, flows between the two FM electrodes. The average thickness of the Al₂O₃ layer is 2.9 nm.

This value is compatible with the effective barrier width fit- ted above by the I – V curve. The average ratio of the com- position of Al–O is 2.00:3.06 from EDS analysis, revealing high insulator quality.

Although the metallic layers are polycrystalline, the in- plane anisotropy has a geometrically induced preferable ori- entation. In the TMR measurement, the applied magnetic field was along the easy axis. Figure 3 shows the typical TMR loop for the SDT junction with the additional 1.6 nm thick CoFe interlayer. This junction exhibits a saturated RA value of 36.7 M⍀␮^m². The magnetization of the soft CoFe/

NiFe layer switches first at a magnetic field of⫺10 Oe 共⫹10 Oe兲 for decreasing 共increasing兲 branch, while the bottom Co electrode switches at a field of⫺14 Oe 共⫹14 Oe兲. The inset in Fig. 3 displays the corresponding MOKE response. The step feature of the hysteresis loop reveals the antiparallel magnetization state over the field range, 10–14 Oe. The maximum TMR ratios are 18.7% and 17.5% for decreasing and increasing branches of TMR loop, respectively, before the magnetization switching of the Co electrode. The highest tunnel resistance during the sweeping of the magnetic field may be missed since the tunnel resistance is very sensitive to

FIG. 1. I – V curve for SDT junction, Co/Al2O3/2.0 nm Co/NiFe.

FIG. 2. Cross-sectional TEM image of SDT junction, Co/Al₂O₃/1.2 nm CoFe/NiFe.

FIG. 3. Typical room temperature TMR loop of SDT junction, Co/Al2O3/1.6 nm CoFe/NiFe. The magnetic field is applied along the easy axis. The inset shows the corresponding hysteresis loop.

6223

J. Appl. Phys., Vol. 90, No. 12, 15 December 2001 Hoet al.

(3)

the magnetic field near the bottom electrode switching field and the magnetic field resolution limit is 0.5 Oe. Missing the highest resistance slightly changes the maximum TMR ratios for decreasing and increasing branches of the TMR loop.

Figure 4 displays a EELS map of an SDT junction with the additional 1.2 nm thick CoFe interlayer. Figure 4共a兲 shows the zero loss map of four different layers. From top to bottom, they are milled substrate, the electrode Co, the insulator Al₂O₃, and the electrode with the additional FM interlayer CoFe/NiFe. This map is similar to the TEM image of Fig. 2. The ultra-thinness of CoFe is such that CoFe and NiFe layers can be distinguished neither by TEM nor the zero loss EELS map. However, as shown in Fig. 4共b兲, the weak but clear Co signal of CoFe layer next to the Al₂O₃ layer can be observed in the Co EELS map at the same place as the signal from Fig. 4共a兲, where the energy loss is fixed at 779 eV of K edge for Co. A complete CoFe layer is grown on the Al₂O₃ layer. Figure 4共b兲 shows the relative EELS inten- sity of the Co line scan profile. The thickness of CoFe is estimated as 1.22 nm.

Figure 5 summarizes the normalized TMR as a function of the additional Co and CoFe interlayer thicknesses. The normalized TMR is defined as the ratio of the TMR ratio of

all SDT junctions, to that of the junction without the additional FM interlayer. For SDT junctions with the additional Co interlayer, the normalized TMR increases from 1 to 2.16 as Co thickness increases from 0 nm to 0.8 nm 共region I兲.

The normalized TMR of the SDT junctions with the additional CoFe interlayer increases from 1 to 3.80 with CoFe interlayer thickness, in a similar region共0–1.0 nm兲. Accord- ing to the Julliere’s model, this result implies that the effective polarization values of SDT junctions are increased by the presence of the additional FM interlayer. As claimed by Slonczewski,⁴ the effective polarization of the tunnel electron is chiefly governed by the ferromagnet-insulator coupling. Thus, the polarization values of the additional FM interlayer near the insulator layer must be an important source of the increased TMR ratio found in SDT junctions with a Co or CoFe interlayer. For SDT junctions with the additional Co and CoFe FM interlayer thickness ranging from the end of region I to 2.0 nm共region II兲, the normalized TMR remains in the range, 2.06 –2.16共8.7%–9.1% of the TMR ratio兲, and 3.80– 4.45 共16.0%–18.7% of the TMR ratio兲, respectively.

Furthermore, in region III, the normalized TMR for both additional Co and CoFe interlayers decreases dramatically to approximately 1, almost equal to the TMR ratio of a junction without the additional FM interlayer. The inset in Fig. 5 presents the field range of the antiparallel magnetization state, H_{A P}, which is defined as the switching field difference between two electrodes, while varing the additional Co and CoFe interlayers. Overall, H_{A P} declines monotonically with increasing thickness of the additional FM interlayer. This result implies that the increasing TMR in region I and the variation observed in region II follow from a change in the transport behavior, rather than from changes in the coercive field which would enable a more stable or extended antiparallel magnetization state.

The polarization value of the FM interlayer can be considered to be the effective polarization value of the FM electrode, since the additional FM interlayer with region II thickness entirely covers Al₂O₃ as shown in Fig. 4共b兲. Previous studies have reported polarization values of CoFe, Co, and NiFe are 47%–53%, 34%– 45%, and 32%– 48%.^10–12 The wide range of polarization values for these three materials may follow from the quality of samples and experimental

FIG. 4. EELS maps of SDT junction共Co/Al2O3/1.2 nm CoFe/NiFe兲. 共a兲 Zero loss map. The tilted substrate is prepared by ion milling;共b兲 Co map.

The energy loss is fixed at 779 eV of K edge for Co. The inset is the line scan 共white straight line across the map兲 for the Co profile. The low Co signal of the additional 1.22 nm CoFe interlayer is due to the limit of the nanobeam probe.

FIG. 5. Normalized TMR of SDT junctions as functions of the additional FM共Co and CoFe兲 interlayer thickness. The normalized TMR is defined as the ratio of the TMR ratio for SDT junctions to that for a junction without the additional FM interlayer 共Co/Al2O3/NiFe兲. The inset shows the field range of the antiparallel magnetization state, HA P, which is defined as the switching field difference between the top and the bottom electrodes, as a function of the additional FM interlayer.

6224 J. Appl. Phys., Vol. 90, No. 12, 15 December 2001 Hoet al.

(4)

methods employed. Possibly the highest normalized TMR can thus be estimated as 1.50 and 1.85 for the additional Co and CoFe interlayers, respectively, according to Julliere’s model. However, these two values are much lower than the experimental results共2.16 and 4.45兲 presented in Fig. 5. An- other influencing factor can be assumed to apply. The various spin-flip scattering factors due to various magnetic ions at the interface of Al₂O₃ and ferromagnetic layers, such as Ni^⫹2, Ni^⫹3, and Co^⫹2 ions,³ may influence the spin- dependent transport behavior, and thus the TMR ratio. The presence of magnetic ions may follow from the diffusion of oxygen ions from the Al₂O₃ layer to FM electrodes, due to the thermal stability and the activated energy difference between Al₂O₃ and FM electrodes.

In region I of Fig. 5, the TMR ratio increases monotonically with increasing additional FM interlayer thickness. Two causes may apply this phenomenon. The first is the cluster- like formation of the additional FM interlayer. Unfortunately, the limited EELS resolution共⬃1 nm兲 is such that the EELS map of the SDT junction in region I cannot be clearly determined. However, the assumption of the discrete ultrathin additional FM interlayer remains reasonable in the common sputtering growth process. Both the cluster-like additional FM interlayer and the NiFe layer contribute to the effective polarization value of the top FM electrode. Accordingly, the effective polarization exceeds that of the pure NiFe layer.

Another possible reason is the size effect of the additional FM interlayer on the polarization value. Upadhyay et al.¹² elucidated the size effect on the polarization value of ultrathin Co films共under 1 nm thick兲, determined in a transport experiment. A highly sensitive and monotonic increase in the polarization with Co thickness was identified. This result suggests that the polarization values of the additional FM interlayers 共both Co and CoFe兲, and in turn, the effective electrode polarization increase rapidly with the thickness of the additional FM interlayer over the low coverage range.

As indicated by the EELS data in Fig. 4共b兲, the additional FM interlayer with a thickness in region II exhibits complete layer formation. The polarization of the FM interlayer may reach a saturation value. Hence, the TMR ratio in region II is maintained almost constant. As mentioned above, the large difference between the results for the additional Co and CoFe interlayers in region II may follow from two possible causes. The first is the polarization behavior. The sec- ond is the presence of the ionized Fe at the interface between the insulator and the top electrode. Jansen et al. demon- strated the enhancement of TMR by Fe ions.⁵

The coercivity of the top FM electrode increases with the additional FM interlayer thickness due to the directive

exchange coupling of the additional FM interlayer and NiFe as the coercivity of the additional FM interlayer exceeds that of NiFe. Thus, the field range of antiparallel spin alignment between the top and bottom FM electrodes decreases with an increasing additional FM interlayer thickness. For region III, the antiparallel field range approaches zero. The absence of a perfect antiparallel spin alignment between the top and bottom FM electrodes reduces the normalized TMR for both the additional Co and the additional CoFe interlayer.

IV. CONCLUSION

The TMR was demonstrated to be strongly influenced by adding the additional FM interlayer at the interface between the insulator layer and the FM electrode. An enhanced factor of 2.16共4.45兲 times the TMR ratio of the SDT junctions was obtained by adding a 0.8 –2.0 nm Co 共1.0–2.0 nm CoFe兲 interlayer at the interface of the insulator and the FM electrode. The EELS map showed generation of a continuous additional FM interlayer in the ultrathin limit at ⬃1.2 nm.

The presence of an ultra-thin additional FM interlayer may change the detailed behavior of the electrode-insulator coupling at interface, possibly leading to a complex interplay between the effective polarization and the spin-flip scattering process, in turn greatly influencing the TMR.

ACKNOWLEDGMENTS

This work was supported by the National Science Coun- cil under Grant No. NSC 89-2119-M-002-014 共M.T.L.兲 and the Topic Project of Academia Sinica共M.T.L.兲 and 共Y.D.Y.兲 in Taiwan.

1W. J. Gallagher, S. S. P. Parkin, Y. Lu, X. P. Bian, A. Marley, K. P. Roche, R. A. Altman, S. A. Rishton, C. Jahnes, T. M. Shaw, and G. Xiao, J. Appl.

Phys. 81, 3741共1997兲.

2M. Julliere, Phys. Lett. A 54, 225共1975兲.

3R. Jansen and J. S. Moodera, J. Appl. Phys. 83, 6682共1998兲.

4J. C. Slonczewski, Phys. Rev. B 39, 6995共1989兲.

5R. Jansen and J. S. Moodera, Appl. Phys. Lett. 75, 400共1999兲.

6L. C. Chen, F. R. Chen, J. J. Kai, L. Chang, J. K. Ho, C. S. Jong, C. C.

Chiu, C. N. Huang, C. Y. Chen, and K. K. Shih, J. Appl. Phys. 86, 3826 共1999兲.

7J. C. Simmons, J. Appl. Phys. 34, 2581共1963兲.

8F. Montaigne, J. Nassar, A. Vaure´s, F. van Dau Fnguyen, F. Petroff, A.

Schuhl, and A. Fert, Appl. Phys. Lett. 73, 2829共1998兲.

9J. J. Sun and P. P. Freitas, J. Appl. Phys. 85, 5264共1999兲.

10J. S. Moodera and G. Mathon, J. Magn. Magn. Mater. 200, 248共1999兲.

11S. K. Upadhyay, R. N. Louie, and R. A. Buhrman, Appl. Phys. Lett. 74, 3881共1999兲.

12B. Nadgorny, R. J. Soulen, Jr., M. S. Osofsky, I. I. Mazin, G. Laprade, R.

J. M. van de Veerdonk, A. A. Smite, S. F. Cheng, E. F. Skelton, and S. B.

Qadri, Phys. Rev. B 61, R3788共2000兲.

6225

J. Appl. Phys., Vol. 90, No. 12, 15 December 2001 Hoet al.