Electric breakdown in ultra-thin MgO tunnel barrier junctions for spin-transfer torque switching
M. Schäfers, V. Drewello, G. Reiss, A. Thomas, K. Thiel, G. Eilers, M. Münzenberg, H. Schuhmann, M. Seibt
EElectric breakdown in ultra-thin MgO tunnel barrier junctions forspin-transfer torque switching
M. Sch¨afers, ∗ V. Drewello, G. Reiss, and A. Thomas
Bielefeld University, Department of Physics,Thin Films and Physics of Nanostructures, 33501 Bielefeld, Germany
K. Thiel
Fraunhofer Institut f¨ur Fertigungstechnik und Angewandte Materialforschung,Wiener Str. 12, 28359 Bremen, Germany
G. Eilers, M. M¨unzenberg, H. Schuhmann, and M. Seibt
I. & IV. Physikalisches Institut and Sonderforschungsbereich 602,Friedrich-Hund-Platz 1, Georg-August-Universit¨at G¨ottingen, 37077 G¨ottingen, Germany (Dated: October 22, 2018)
Abstract
Magnetic tunnel junctions for spin-transfer torque switching were prepared to investigate thedielectric breakdown. The breakdown occurs typically at voltages not much higher than the switch-ing voltages, a bottleneck for the implementation of spin-transfer torque Magnetic Random AccessMemory. Intact and broken tunnel junctions are characterized by transport measurements andthen prepared for transmission electron microscopy and energy dispersive x-ray spectrometry byfocussed ion beam. The comparison to our previous model of the electric breakdown for thickerMgO tunnel barriers reveals significant differences arising from the high current densities.
PACS numbers: 68.37.Lp, 85.30.Mn, 85.75.-d ∗ Electronic address: [email protected]; URL: a r X i v : . [ c ond - m a t . m t r l - s c i ] J u l he switching of magnetic tunnel junctions (MTJs) by spin-transfer torque has gainedhigh interest in the last years [1, 2, 3]. From a technological point of view this effect canbe used for compact microwave oscillators [4] or a new type of magnetic random accessmemory (MRAM) [5, 6], which does not need special writing lines above the MTJ cells.This spin-transfer torque (STT) MRAM can be integrated at much higher densities whichis essential for applications.As high current densities are required to switch MTJs with the STT method, MTJs witha thin barrier and, therefore, low area resistance are used. Still a relatively high voltageis applied to the junction to achieve the switching currents. These are in the range of afew hundred mV compared to only a few mV needed for normal (read) operation [1]. Theapplication of higher voltages is limited, as at some point an electrical breakdown of thebarrier is observed [7]. Furthermore, this breakdown happens at lower voltages for thinnerbarriers [8].The breakdown voltage should be much higher than the switching voltage in order tooptimize the stability of the junctions. The breakdown effect has been widely investigated byelectrical transport measurements [8, 9, 10, 11]. While this makes an optimization with largeamounts of samples more practicable it provides little information about the microscopicprocesses during or after a breakdown.In this paper we present transmission electron microscopy analysis of MTJs which werestressed through a dielectric breakdown. They show a very different breakdown behaviorcompared to thicker barriers [12].The breakdown investigated in this study is commonly called ’hard breakdown’. A suddenincrease in the current (during constant voltage stress in this case) indicates this type ofbreakdown.The MTJs are sputter deposited in a Singulus ndt timaris ii cluster tool. Thelayer stack is 5 (nm) Ta/90 Cu-N/5 Ta/20 Pt-Mn/2.2 Co Fe /0.8 Ru/2 Co Fe B /1.1MgO/1.5 Co Fe B /10 Ta/30 Cu-N/7 Ru. The complete stack is annealed for 90 minutesat 360 ◦ C in a magnetic field of 1 T.Elliptic junctions sized 360 nm ×
150 nm are patterned into a resist layer by conventionalelectron beam lithography. Argon ion beam etching transfers the patterns into the layerstack. The complete sample is covered with a thick layer of tantalum oxide to electricallyinsulate the separated pillars from each other. The top of the pillars are opened with a2 r e s i s t a n c e ( ! )
75 75-33 320magnetic field (Oe) 0-0.5 0.50.32-0.24pulse voltage (V)
FIG. 1: Magnetic minor loop and resistance vs. voltage plot of the magnetic tunnel junction shownin the left images of Fig. 3 and Fig. 4. lift-off process. A gold layer is deposited as upper lead material. Contact pads are createdby a second exposure and etching process.The samples for high-resolution electron microscopy (HRTEM) investigation have beenprepared by focused ion beam (FIB) with a FEI Nova Nanolab 600 instrument. Damage tothe device caused by the Ga + ions during milling was prevented by a 0.5 µ m Pt coating ontop of the lamella. A last low energy etching step at 5 kV under an incidence angle of 7 ◦ wasperformed to reduce the amorphous surface layer resulting from the cut at 30 kV Ga + beam.The final thickness of the lamella is in the range of 10-20 nm. TEM work was done using aPhilips CM200-FEG UT operated at an accelerating voltage of 200kV. The microscope hasa point resolution of 0.19 nm and an information limit of 0.11 nm. Energy-dispersive x-rayspectrometry (EDX) was performed using a Si:Li detector (Link ISIS).About 200 magnetic tunnel junctions were prepared by e-beam lithography and examinedby conventional 2-terminal transport investigations to determine the characteristics of thejunctions. Figure 1 (left) shows the magnetic minor loop of one of the junctions. A TMRratio of 97% was reached. This is a typical value for low resistive tunnel barriers for STTswitching [13, 14].The corresponding STT loop is shown on the right hand side of Figure 1. The TMR ratiois now 92% due to slightly changed resistances in the parallel and anti-parallel states. Thisis caused by diffusion processes and explained for these junctions in detail by Krzysteczkoet al. [15] The switching voltages of − .
24 V and 0 .
32 V correspond to switching current3 r e s i s t a n c e ( ! ) -75 -9 32 75magnetic field (Oe) FIG. 2: Magnetic minor loop and resistance vs. voltage plot of the MTJ shown in the right imagesof Fig. 3 and Fig. 4 before it was stressed through the dielectric breakdown. densities of − . × A / cm and 3 . × A / cm , respectively.Figure 2 depicts the magnetic minor and STT loops of another magnetic tunnel junction.Here, the TMR ratio is 86% for the magnetic loop and 82% in the STT case. The currentdensities are − . × A / cm switching from parallel to anti-parallel and 3 . × A / cm viceversa. A few junctions were stressed through a dielectric breakdown after the measurementsincluding the junction investigated in Figure 2.The left hand side of Figure 3 shows a TEM image of the junction characterized in Figure1. The complete layer stack excluding the lower Ta layer is visible. The Ta oxide on theleft and right hand side of the image and the upper lead are necessary for the preparationof the junction by lithography. The junction width as seen in the TEM data depends onthe position where the ellipsis was cut by the focussed ion beam. The thin bright whiteline just underneath the Ta oxide is the MgO tunnel barrier. It extends through the fullsize of the image. We verified the smoothness of the three dimensional extension of theMgO barrier into the of 10-20 nm thick lamella by transmission electron tomography. SeeEPAPS supplementary material at [URL will be inserted by AIP] for a movie of transmissionelectron tomography reconstructed from TEM images.The same target preparation was applied to the stressed junction characterized in Fig-ure 2. The TEM image of this junction is depicted on the right hand side of Figure 3. AllTEM images and EDX scans are rotated and scaled to reveal the same region of the layerstack for a better comparison. 4 aupper leadPt-MnTaoxide Cu-NPt50 nm Co-Fe-BCo-Fe-BRuMgOCo-Fe ]Ru TaoxideCu-N135 nmTaupper leadCu-NPt-MnTaoxide Cu-NPt50 nm 180 nm
Co-Fe-BCo-Fe-BRuMgOCo-Fe ]Ru Taoxide brokenintact
FIG. 3: Transmission electron microscopy image of an intact (left) and a broken (right) magnetictunnel junction. The Ta oxide and the upper lead are necessary for the definition of the junctionby the lithography processes.
The most dramatic change is seen in the region of the upper Cu-N. Within the brightregions, all material has been competely removed during breakdown. A closed up look tothe tunnel barrier region reveals that the thin white line, the signature of the MgO barrier,can not be identified on a width of about 80 nm. One concludes that the barrier within thatregion has completely vanished.To solidify this conclusion drawn from the TEM images, we carried out energy dispersivex-ray (EDX) investigations of the TEM slices to yield element specific images. The leftimage in Figure 4 shows the intact junction. Only Cu, Co-Fe and Ta are mapped here.The lateral resolution of the EDX is limited, therefore, the ferromagnetic electrodes are notseparated from the tunnel barrier, but visible as a thick red stripe.The extent of the damage after breakdown is obvious in the EDX map of the tunnel barrierregion and the adjacent layers in the right image of Figure 4. Not only the tunnel barrier butthe ferromagnetic electrodes are affected by the high current density of 7 × A / cm duringand 9 × A / cm after breakdown. Within the missing barrier, the material is completlydisplaced by the local heating and electrostatic forces (”electron wind”) arising from the e − -flux from the bottom to the top originating from the high current densities. The electrodematerial is found 40 nm above the position of the element (Figure 4, right). We comparedthe different junction types in table I including alumina based junctions [7, 16] and thicker(2.0 nm) MgO based systems [12].Even though the mechanisms of the electric breakdown are similar, the consequences5 aupper leadCu-NPt-MnTaoxide Cu-NPt50 nm 180 nm ]Ru Taoxide Taupper leadPt-MnTaoxide Cu-NPt50 nm ]Ru TaoxideCu-N135 nm Cu Co-Fe Ta Cu Co-Fe Ta
FIG. 4: EDX scan across the TEM image of an intact (left) and a broken (right) magnetic tunneljunction overlaid to the transmission electron microscopy image of an intact (left) and a broken(right) magnetic tunnel junction. Cu, Co-Fe and Ta distributions are mapped in the image.barrier material AlO x MgO MgObarrier thickness nm 1.8 2.0 1.1junction area µ m
10k 90 0.044area resistance at 10 mV Ω µ m
10M 50k 8breakdown (bd) voltage V 1.5 1.6 0.4resistance at bd Ω ∼
600 330 130resistance after bd Ω ∼ ∼ ∼ / cm
24 5k 7Mcurrent density after bd A / cm ∼ ∼ ∼ arising from the different area resistances are very different. Single pinholes were reportedin the case of alumina. We showed in our earlier work that multiple pinholes were foundin the case of the thicker MgO [12] and explained the effect by modeling the local currentdistribution. Local heating resulted in a crystallization of the electrodes above the pinholes.The pinhole distance was found to be about 50 to 100 nm. Here the structural confinementdoes not allow to form multiple pinholes in the element. The numerous pinholes merge and6orm a large break in the tunnel barrier. Furthermore drastic effects due to the increasedcurrent density of 9 × A / cm (MgO 1.1 nm MgO) as compared to 1 . × A / cm (2.0 nmMgO) have three possible consequences: The temperature increase is large enough to eitherheat the element significantly above the Co-Fe-B crystallization point or even local melting.Also, the strong local electrostatic forces at the barrier that act on the TMR stack maylead to electromigration. The e − -flux direction from the bottom to the top determinesthe direction of the material transport. The relative contributions of these processes tothe microstructural changes can not be deduced from the present study and need furtherinvestigations.In summary, we presented electric characterizations and TEM images with EDX scans ofmagnetic tunnel junctions showing current induced magnetization switching. The junctionswith ultra-thin tunnel barriers were investigated before and after a dielectric breakdown.We observed a breakdown of the tunnel barrier over several 10 nm and damage of the adja-cent ferromagnetic electrodes. It can be explained by the large current density and lateralconfinement in case of the spin-transfer toque devices. Acknowledgments
We acknowledge J. Schmalhorst for helpful discussions and Singulus Nano DepositionTechnology for providing the layer stacks. Support by the Deutsche Forschungsgemeinschaftwithin the priority program SFB 602 and the research grant [1] Y. Huai, F. Albert, P. Nguyen, M. Pakala, and T. Valet, Appl. Phys. Lett. , 3118 (2004).[2] G. D. Fuchs, N. C. Emley, I. N. Krivorotov, P. M. Braganca, E. M. Ryan, S. I. Kiselev, J. C.Sankey, D. C. Ralph, R. A. Buhrman, and J. A. Katine, Appl. Phys. Lett. , 1205 (2004).[3] J. Hayakawa, S. Ikeda, Y. M. Lee, R. Sasaki, T. Meguro, F. Matsukura, H. Takahashi, andH. Ohno, Jpn. J. Appl. Phys. , L1267 (2005).[4] A. M. Deac, A. Fukushima, H. Kubota, H. Maehara, Y. Suzuki, S. Yuasa, Y. Nagamine,K. Tsunekawa, D. D. Djayaprawira, and N. Watanabe, Nature Phys. , 803 (2008).
5] M. Hosomi, H. Yamagishi, T. Yamamoto, K. Bessho, Y. Higo, K. Yamane, H. Yamada,M. Shoji, H. Hachino, C. Fukumoto, et al., IEDM Tech. Dig. p. 459 (2005).[6] J. A. Katine and E. E. Fullerton, J. Magn. Magn. Mater. , 1217 (2008).[7] W. Oepts, H. J. Verhagen, W. J. M. de Jonge, and R. Coehoorn, Appl. Phys. Lett. , 2363(1998).[8] A. A. Khan, J. Schmalhorst, A. Thomas, O. Schebaum, and G. Reiss, J. Appl. Phys. ,123705 (2008).[9] W. Oepts, M. F. Gillies, R. Coehoorn, R. J. M. van de Veerdonk, and W. J. M. de Jonge, J.Appl. Phys. , 8038 (2001).[10] B. Oliver, Q. He, X. Tang, and J. Nowak, J. Appl. Phys. , 4348 (2002).[11] J. Schmalhorst, H. Br¨uckl, M. Justus, A. Thomas, G. Reiss, M. Vieth, G. Gieres, andJ. Wecker, J. Appl. Phys. , 586 (2001).[12] A. Thomas, V. Drewello, M. Schaefers, A. Weddemann, G. Reiss, G. Eilers, M. Muenzenberg,K. Thiel, and M. Seibt, Appl. Phys. Lett. , 152508 (2008).[13] H. Kubota, A. Fukushima, Y. Ootani, S. Yuasa, K. Ando, H. Maehara, K. Tsunekawa, D. D.Djayaprawira, N. Watanabe, and Y. Suzuki, Jpn. J. Appl. Phys. , L1237 (2005).[14] S. Serrano-Guisan, K. Rott, G. Reiss, and J. Langer, Phys. Rev. Lett. , 087201 (2008).[15] P. Krzysteczko, X. Kou, K. Rott, A. Thomas, and G. Reiss, J. Magn. Magn. Mater. , 144(2009).[16] K. Shimazawa, N. Kasahara, J. Sun, S. Araki, H. Morita, and M. Matsuzaki, J. Appl. Phys.(2000)., 144(2009).[16] K. Shimazawa, N. Kasahara, J. Sun, S. Araki, H. Morita, and M. Matsuzaki, J. Appl. Phys.(2000).