Vol.2 No.3 2009

Research paper : Creating non-volatile electronics with spintronics technology (S. Yuasa et al.)−198−Synthesiology - English edition Vol.2 No.3 (2009) conventional amorphous Al-O tunnel barrier and the crystalline MgO(001) tunnel barrier is illustrated in Fig. 7. Conduction electron states that have various wave function symmetries (Bloch states) exist in the ferromagnetic electrode. For an amorphous Al-O tunnel barrier, the symmetry of atomic arrangement in the barrier is broken, so the various Bloch states are mixed up within the electrode and tunneling conduction occurs (Fig. 7(a)). Each Bloch state in the electrode generates a positive or negative MR ratio, depending on its orbital symmetry. With an amorphous Al-O barrier, those Bloch states are mixed and all are subject to tunneling conduction. Therefore, the mean MR ratio of the Bloch states (which is to say the MR ratio of the MTJ device) cannot attain a high value, and an MR ratio that greatly exceeds 70 % cannot be attained at room temperature.If the tunnel barrier is crystalline MgO (001), on the other hand, an entirely different characteristic is predicted by theory. The single-crystal MTJ device tunneling process is illustrated in Fig. 7(b). The tunnel electron is often assumed to be a free electron, but the evanescent states of electrons that are in the actual band gap of the insulating tunnel barrier have a special orbital symmetry and a special band dispersion that differ greatly from free electrons. Three kinds of evanescent states exist within the MgO (001) band gap: 1 (spd hybridized high symmetry state), 5 (pd mixed state), and 2’ (d low electronical symmetry state). The attenuation rates of the density of states of these states within the tunnel barrier depend greatly on the orbital symmetry of the state. The 1 evanescent states have the slowest attenuation in the tunnel barrier (i.e., the longest attenuation length). Accordingly, the tunneling current via this 1 states is dominant (Fig. 7(b)). In an ideal tunneling process, only the 1 Bloch states in the Fe (001) electrode can couple with the 1 evanescent states in the MgO, so the dominant tunneling path is Fe-1 MgO-1Fe-1. What should be noted here is that the 1 Bloch state in the Fe (001) electrode is a special electron state that can generate a very large positive MR ratio. From the fact that only electrons that have the 1 symmetry selectively pass through the MgO tunnel barrier, as we see in Fig. 7(b), we can theoretically expect a huge MR ratio of over 1000 %. The theoretical prediction for such a huge MR ratio is not limited to the bcc Fe (001) electrode, but is predicted also for Fe and Co based ferromagnetic alloys that have the bcc structure.2.2 Achieving the MgO tunnel barrier giant TMR effectWhen the theoretical prediction of the giant TMR effect for a crystalline MgO tunnel barrier appeared in 2001, there were experimental attempts to actually fabricate MTJ devices that had the single-crystal Fe (001)/MgO (001)/Fe (001) structure, mainly by public research organizations in Europe, but there was no success. Room temperature MR ratios that exceeded those of conventional amorphous Al-O tunnel barrier were not attained, so the expectations for the crystalline MgO tunnel barrier were not met and the theoretical predictions for the giant TMR effect came to be viewed with skepticism. Under those circumstances, AIST continued experimental research on crystalline MgO tunnel barriers, and succeeded in fabricating a high-quality single-crystal Fe (001)/MgO (001)/Fe (001) MTJ device using molecular beam epitaxy (MBE) in 2004 (Fig. 8)[7][8]. That single-crystal MgO-MTJ device was used to achieve the world’s first room temperature MR ratio that exceeded that of an amorphous Al-O barrier in the beginning of 2004 (Fig. 9, 1))[7]. That paper also verified high reproducibility, excellent voltage characteristics and other aspects of practicality, and was thus a historical turning point that brought the crystalline MgO tunnel barrier back into the limelight. After that, AIST achieved an even higher room temperature MR ratio of 180 % by further improving the quality of the crystalline MgO tunnel barrier in the latter half of 2004 (Fig. 9, 2))[8]. On the other hand, at about the same time that AIST produced those results, Parkin et al. of IBM fabricated an MTJ device that used a preferred-oriented polycrystalline (textured) MgO (001) tunnel barrier with Fig. 6 Non-volatile memory MRAM (a) Schematic diagram, (b) Circuit diagram, (c) Cross-sectionFig. 7 Electron tunneling transport. (a) Amorphous Al−O tunnel barrier, (b) crystalline MgO(001) tunnel barrier (a)(b)(c)orBit lineCMOSWard lineWard lineMTJ deviceCMOSn+p n+ Ward lineBit lineMTJ deviceorMTJ deviceBit lineWrite lineP state“0”AP state“1” Δ1Δ1Δ1Δ5Δ2 Fe(001)Fe(001)Δ1 Δ5Δ2 Fe(001)(a)CrystallineMgO(001)AmorphousAl-O(b)


page 25