High temperature annealing induced superparamagnetism in CoFeB ...

JOURNAL OF APPLIED PHYSICS 108, 083901 共2010兲

High temperature annealing induced superparamagnetism in CoFeB/MgO/CoFeB tunneling junctions Xiaoming Kou,1,a兲 Weigang Wang,2 Xin Fan,1 Lubna R. Shah,1 Rae Tao,1 and John Q. Xiao1,b兲 1

Department of Physics and Astronomy, University of Delaware, Newark, Delaware 19716, USA Department of Physics and Astronomy, Johns Hopkins University, Baltimore, Maryland 21218, USA

2

共Received 1 February 2010; accepted 31 August 2010; published online 18 October 2010兲 We have investigated the evolution of the magnetic transport properties as a function of short annealing time in CoFeB/MgO/CoFeB based magnetic tunnel junctions 共MTJs兲 with a free layer of 2 nm. It is found that the hysteresis behaviors in magnetoresistance 共MR兲 loops disappear in samples annealed for 17 min. The linear region between MR and the applied field gradually increases. The MR loops without hysteresis can be well fitted by using the superparamagnetism theory, suggesting the formation of superparamagnetic particles in the free layer during the high temperature annealing. The control of MTJ properties with annealing time is desirable in magnetic field sensor productions. © 2010 American Institute of Physics. 关doi:10.1063/1.3496663兴 I. INTRODUCTION

The tunneling magnetoresistance 共TMR兲 in MgO-based magnetic tunnel junctions 共MTJs兲 has been extensively studied since 2004.1,2 CoFeB electrodes are widely used because they not only promote the growth of crystalline 共100兲 MgO barrier, but also have required half-metal-like band structures for high TMR.3,4 Besides various applications such as read head,5 microwave oscillators,6 magnetic random access memory,7 pressure sensors,8 and microwave detectors,9 a great deal of attention has been focused on its application in magnetic field sensors. The sensor application requires a hysteresis free MR loop and a linear response of TMR with respect to the magnetic field. There are two strategies to achieve these features. One can use a perpendicular magnetization configuration between two magnetic electrodes of rectangular or ellipsoidal shape with large aspect ratio.10–12 Alternatively, one can introduce the superparamagnetic behavior in the free layer by reducing its thickness.13–16 The latter as the advantage of a simplified fabrication process, although a precise control of the free layer thickness is necessary. In both techniques, high temperature annealing is an indispensable step to achieving a high TMR value in MgObased junctions. The annealing is usually hours long, taking place in a strong magnetic field and a high vacuum. Recent research17,18 revealed that a high TMR can quickly develop at the early stage of annealing in air or Ar, so the timeconsuming annealing and expensive vacuum equipment become unnecessary. In this study, we fabricated CoFeB/MgO/ CoFeB MTJ junctions with a relatively thin free layer and studied the evolution of the MR loops over annealing times at high temperature in Ar. We found that the MR loop becomes almost hysteresis free after 17 min of annealing and that the linear region gradually increases with annealing time and a tradeoff in TMR value. The hysteresis free loops can a兲

Electronic mail: [email protected]. Electronic mail: [email protected].

b兲

0021-8979/2010/108共8兲/083901/4/$30.00

be well fitted by using the theory for superparamagnetism. This implies the formation of superparamagnetic particles in the free layer during high temperature annealing. II. EXPERIMENT

The MTJ samples were deposited in a magnetron sputtering system with a base pressure of 8 ⫻ 10−8 Torr. The structure was Si/ SiO2 / Ta 7/Ru 20/Ta 7/CoFe 4/IrMn 15/ CoFe 2/Ru 1.7/CoFeB 4/MgO 1.5–3/CoFeB 2/Ta 8/Ru 10, where the numbers are the layer thickness in nanometers. A thick Ru layer of 1.7 nm was used in the synthetic antiferromagnetic structure for better stability at high annealing temperatures.17 A combinational technique was used to form a wedge-shaped MgO barrier layer on the wafer. MTJ junctions with diameters from 25 to 100 ␮m were defined by standard photolithography and ion-beam etching. The annealing was performed in Ar atmosphere on a specially designed sample stage allowing for annealing temperatures up to 500 ° C. The oxygen level is lower than 5 ppm and the H2O level below 0.5 ppm. After being annealed for a certain time duration, the sample was cooled down to room temperature under a magnetic field of 2 kOe to establish the exchange pinning. Subsequently, the TMR ratio of the junctions was measured at room temperature in air by a conventional four-probe method. The same annealing and TMR measurement procedures were repeated several times on the sample in order to study its evolution over annealing time. The magnetic property of the sample was measured by a vibrating sample magnetometer 共VSM, Lakeshore 7404兲. III. RESULTS AND DISCUSSION

Figure 1 illustrates the TMR ratio and coercivity as a function of annealing time for a single MTJ junction, which was annealed at 460 ° C. The TMR ratio in this manuscript is calculated using 共RAP − R P兲 / R P, where R P is the resistance of the parallel state, in which the magnetizations of the free layer and the pinned layer are parallel. RAP is the resistance

108, 083901-1

© 2010 American Institute of Physics

Downloaded 14 Aug 2011 to 128.220.144.249. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions

083901-2

Kou et al.

FIG. 1. 共Color online兲 TMR ratio and coercivity as a function of annealing time for a single MTJ junction annealed at 460 ° C.

in the antiparallel state. The TMR for the as-prepared MTJs were only 5%–8%. The TMR quickly developed to 128% with only 1 min of annealing, similar to other reported results.17 Interestingly, different from reported results, the TMR ratio gradually decreases from 128% to 17% after 70 min of annealing. The coercivity also shows a very different behavior. Little change in coercivity was seen in the first 4 min of annealing. Then the coercivity decreases rapidly from about 17.2 to 2.7 Oe in the next 12 min of annealing. The coercivity remains around 2 Oe for the rest of the annealing. We attribute this unique property to the formation of superparamagnetic particles in the free layer during the annealing, which will be explained in detail later. More than 20 junctions from the same wafer with different barrier thicknesses have been measured and they all show similar results. Selected MR loops of the same junction at different annealing times are depicted in Fig. 2. As shown, the hysteretic

J. Appl. Phys. 108, 083901 共2010兲

FIG. 3. 共Color online兲 The linear region in MR loops as a function of annealing time. The TMR ratios are also labeled.

behavior nearly disappears in samples annealed for longer than 17 min. The linear region gradually expands upon annealing. The dependence of the linear region on the annealing time is displayed in Fig. 3, with the decreasing TMR ratio. The linear region at 70 min ranges from ⫺47 to 68 Oe, more than ten times larger than that at 17 min, from ⫺3.1 to 5.3 Oe. The growth of the linear region accompanied by the decreasing TMR value was also observed in other studies by systematically reducing the free layer thickness.14 The advantage in our sample is that, in sensor fabrication, controlling of annealing time is much easier than tuning the free layer thickness in the range between 0.8 to 1.5 nm. In order to explore the origin of the MR loops without hysteresis, blank samples without patterning were annealed for different durations at 460 ° C and measured by the VSM. The magnetic field was applied in plane, perpendicular to the pinning direction. The results indicate that no easy axis in the free layer was induced in the direction perpendicular to the

FIG. 2. 共Color online兲 Measured MR loops of a MTJ junction annealed for different times at 460 ° C.


083901-3

J. Appl. Phys. 108, 083901 共2010兲

Kou et al.

field. After rearranging the equation, the change in the conductance becomes ⌬G = G − G0 ⬀ cos共␪F − ␪ P兲. The theoretical calculation13 shows that the superparamagnetic particles in the free layer have a pancake shape. Thus, it is reasonable to assume that the magnetic moments stay in the film plane due to the shape anisotropy. Therefore we simplify the Langevin equation into a two-dimensional case. For an assembly of paramagnetic particles, the cos共␪F − ␪ P兲 can be written as ⌬G ⬀ cos共␪F − ␪ P兲 =

FIG. 4. 共Color online兲 Measured MR loops of an MTJ junction annealed for 20 min at 460 ° C, ␸ is the angle between the applied magnetic field and the pinning direction. The solid line is the fitting based on the superparamagnetism theory.

pinning field by annealing. This rules out the possibility of a perpendicular magnetization configuration between two ferromagnetic electrodes formed during annealing, which may eliminate the hysteretic behavior. Another MTJ sample with the same structure but a thicker free layer of 3 nm was fabricated. Annealed at the same temperature, the sample shows 200% MR with little change in the first 15 min of annealing. After that, the TMR decreases and the coercivity remains at about 17 Oe. The reduction in the TMR can be attributed to the Mn diffusion during the high temperature annealing.19 The same mechanism may also contribute to the TMR decrease in the sample with a thin free layer of 2 nm. However, it cannot explain the sudden change in coercivity. This comparison suggests the thickness of the CoFeB free layer is important to the unique property observed here. In the aforementioned strategy of reducing the free layer thickness, a critical thickness exists, below which the free layer becomes superparamagnetic. The critical thickness is usually about 1.4 nm, smaller than the 2 nm in our junctions. Nevertheless, the annealing temperature of 460 ° C is much higher than the temperatures used by other groups. It is reasonable to consider that the high thermal energy changes the critical thickness, leading to superparamagnetic behavior in a relatively thick free layer. The superparamagnetism theory is used to study the MR loops of the junction with 2 nm thick free layer. The junction was annealed for 20 min at 460 ° C. The MR loops were measured with different angles between the applied field and the pinning direction, illustrated in Fig. 4. The applied field was in the film plane. For MTJ, the conductance of a junction can be expressed as13 G = G0关1 + P2 cos共␪F − ␪ P兲兴, where P is a constant, and G0 is the conductance when the magnetizations of the free layer and pinning layer are perpendicular to each other. ␪F 共␪ P兲 is the angle between the magnetization of the free 共pinning兲 layer and the applied

兰20␲cos共␪F − ␪ P兲e共HM sV cos ␪F/kT兲d␪F 兰20␲e共HM sV cos ␪F/kT兲d␪F

,

where M s, k, and T are the saturation magnetization 共860 emu/cc兲 of CoFeB,13 the Boltzmann constant, and the temperature 共300 K兲, respectively. H is the applied field. The particle volume V is used as a fitting parameter, which is set to be 1100 nm−3. Another fitting parameter, the pinning field 共240 Oe兲 is used for the loops measured at ␸ = ␲ / 6, ␲ / 3, and ␲ / 2. The fitted results are also shown in Fig. 4, which match the experimental results very well. Although the argument about the critical thickness altered by the high thermal energy is tentative and the underlying mechanism has yet to be elucidated, the excellent fitting implies that the formation of superparamagnetic particles during the annealing is highly plausible. The same analysis was applied to the MR loops annealed for different durations with ␸ = 0. It was found that the fitting parameter, particle volume V, quickly decreases with the annealing time, which seems counterintuitive. It has been known that the boron diffuses during the crystallization of the CoFeB layer.20 One scenario to explain the decreasing particle size is that the initial continuous free layer keeps breaking down into small pieces during annealing because of the boron diffusion. Electronic microscopic study is under way in order to clarify this problem. It should be mentioned that we also annealed the 2 nm free layer sample at 420 ° C. The time dependent study shows a slow decay of TMR from 180% to 148% in 60 min and no coercivity changes, suggesting the high annealing temperature is crucial for the observed superparamagnetic behaviors. The formation of the superparamagnetic particles and the decrease in the particle size upon annealing may cause the magnetic moments in the free layer to be more difficult to be aligned. Therefore, as long as the same measurement fields are used 共⫾150 Oe兲, the parallel conductance G P decreases with annealing while the antiparallel conductance GAP increases, leading to the reduction in the MR value. The evolution of the parallel and antiparallel conductance of the same MTJ junction used above as a function of the annealing time is illustrated in Fig. 5. The conductance in both configurations reduces with the annealing time. The decrease in the conductance was also observed in the previous study,21 where the decrease in G P has been attributed to the impurities diffusion during annealing. However, in the present case the antiparallel conductance decreases much slower than the parallel conductance, supporting the above argument that magnetic moments in the free layer become more difficult to


083901-4

J. Appl. Phys. 108, 083901 共2010兲

Kou et al.

ACKNOWLEDGMENTS

The characterization and analyses related work was supported by DOE under Grant No. DEFG02-07ER46374 and the sampling fabrication was supported by NSF under Grant No. DMR0827249. 1

S. S. P. Parkin, C. Kaiser, A. Panchula, P. M. Rice, B. Hughes, M. Samant, and S. H. Yang, Nature Mater. 3, 862 共2004兲. S. Yuasa, T. Nagahama, A. Fukushima, Y. Suzuki, and K. Ando, Nature Mater. 3, 868 共2004兲. 3 D. D. Djayaprawira, K. Tsunekawa, M. Nagai, H. Maehara, S. Yamagata, N. Watanabe, S. Yuasa, Y. Suzuki, and K. Ando, Appl. Phys. Lett. 86, 092502 共2005兲. 4 W. H. Butler, Sci. Technol. Adv. Mater. 9, 014106 共2008兲. 5 K. Tsunekawa, D. D. Djayaprawira, M. Nagai, H. Maehara, S. Yamagata, N. Watanabe, S. Yuasa, Y. Suzuki, and K. Ando, Appl. Phys. Lett. 87, 072503 共2005兲. 6 A. M. Deac, A. Fukushima, H. Kubota, H. Maehara, Y. Suzuki, S. Yuasa, Y. Nagamine, K. Tsunekawa, D. D. Djayaprawira, and N. Watanabe, Nat. Phys. 4, 803 共2008兲. 7 S. Yuasa, A. Fukushima, T. Nagahama, K. Ando, and Y. Suzuki, Jpn. J. Appl. Phys., Part 2 43, L588 共2004兲. 8 D. Meyners, T. von Hofe, M. Vieth, M. Ruhrig, S. Schmitt, and E. Quandt, J. Appl. Phys. 105, 07C914 共2009兲. 9 X. Fan, R. Cao, T. Moriyama, W. Wang, H. W. Zhang, and J. Q. Xiao, Appl. Phys. Lett. 95, 122501 共2009兲. 10 D. Mazumdar, W. F. Shen, X. Y. Liu, B. D. Schrag, M. Carter, and G. Xiao, J. Appl. Phys. 103, 113911 共2008兲. 11 Y. Lu, R. A. Altman, A. Marley, S. A. Rishton, P. L. Trouilloud, G. Xiao, W. J. Gallagher, and S. S. P. Parkin, Appl. Phys. Lett. 70, 2610 共1997兲. 12 D. Lacour, H. Jaffres, F. N. V. Dau, F. Petroff, A. Vaures, and J. Humbert, J. Appl. Phys. 91, 4655 共2002兲. 13 W. F. Shen, B. D. Schrag, A. Girdhar, M. J. Carter, H. Sang, and G. Xiao, Phys. Rev. B 79, 014418 共2009兲. 14 Y. Jang, C. Nam, J. Y. Kim, B. K. Cho, Y. J. Cho, and T. W. Kim, Appl. Phys. Lett. 89, 163119 共2006兲. 15 P. Wisniowski, J. M. Almeida, S. Cardoso, N. P. Barradas, and P. P. Freitas, J. Appl. Phys. 103, 07A910 共2008兲. 16 J. M. Almeida, P. Wisniowski, and P. P. Freitas, J. Appl. Phys. 103, 07E922 共2008兲. 17 W. G. Wang, C. Ni, A. Rumaiz, Y. Wang, X. Fan, T. Moriyama, R. Cao, Q. Y. Wen, H. W. Zhang, and J. Q. Xiao, Appl. Phys. Lett. 92, 152501 共2008兲. 18 W. G. Wang, J. Jordan-sweet, G. X. Miao, C. Ni, A. K. Rumaiz, L. R. Shah, X. Fan, P. Parsons, R. Stearrett, E. R. Nowak, J. S. Moodera, and J. Q. Xiao, Appl. Phys. Lett. 95, 242501 共2009兲. 19 J. Hayakawa, S. Ikeda, Y. M. Lee, F. Matsukura, and H. Ohno, Appl. Phys. Lett. 89, 232510 共2006兲. 20 J. J. Cha, J. C. Read, W. F. Egelhoff, P. Y. Huang, H. W. Tseng, Y. Li, R. A. Buhrman, and D. A. Muller, Appl. Phys. Lett. 95, 032506 共2009兲. 21 W. G. Wang, C. Ni, G. X. Miao, C. Weiland, L. R. Shah, X. Fan, P. Parson, J. Jordan-sweet, X. M. Kou, Y. P. Zhang, R. Stearrett, E. R. Nowak, R. Opila, J. S. Moodera, and J. Q. Xiao, Phys. Rev. B 81, 144406 共2010兲. 2

FIG. 5. 共Color online兲 Conductances in parallel and antiparallel states as a function of annealing time for the MTJ used in Fig. 1 annealed at 460 ° C. The inserted picture shows the I–V curves of another MTJ junction annealed at the same temperature for 1 and 40 min.

be aligned with annealing, indicating the reduction in TMR is not due to the change in barrier/electrode interface properties caused by annealing. The I–V curves of another MTJ annealed at 460 ° C with different durations are shown in the insert of Fig. 5. The I–V curves of parallel and antiparallel states almost overlap each other after long annealing due to the superparamagnetic nature of the top electrode. IV. CONCLUSION

In summary, we have studied the evolution of the magnetic transport properties as a function of the annealing time in CoFeB/MgO/CoFeB based MTJ junctions with a free layer of 2 nm. It is found that the hysteresis behaviors in MR loops disappear in samples annealed for 17 min. The linear region between MR and the applied field gradually increases. The MR loops without hysteresis can be well fitted by using the superparamagnetism theory, suggesting that superparamagnetic particles form in the free layer during the high temperature annealing. Controlling the MTJ property by altering annealing time is desirable in magnetic field sensor productions.
