Chapter 9
Summary

Among the main results of the thesis are:

1. Introduction and development of a realistic stochastic model of the resistive switching. The dominance of stochastic processes in the RRAM operation degrades the reliability of this memory. This makes RRAM an unlikely candidate for universal memory development, at least in its current stage.
2. Development of methods for achieving a symmetrical switching in asymmetric penta-layers MTJs structures.
3. Introduction of a new concept of the free layer in the form of an in-plane composite free layer and its subsequent structural optimization. The discovered mechanism of switching acceleration and critical current reduction in STT-MRAM is general and does not depend on the magnetic material used for the free layer fabrication. This makes the proposed concepts of STT-MRAM with a composite free layer (C-MTJ and C2-MTJ) a good candidate for the development of future universal memory.
4. Theoretical prediction of a new mechanism for switching failure in a MTJ-based STT-MRAM through formation of the vortex/transverse domain wall in a free layer.
5. Introduction of two new concepts of a bias-field-free spin-torque oscillator.

(1): The first full stochastic microscopic model of oxide resistive switching was developed. The main advantage of the stochastic model is that for the first time the movement of oxygen ions is taken into account. The distribution of the electron occupation probabilities calculated with the model is in good agreement with previous work. It has been shown that the process of rupture of the conductive filament is determined by the dynamics of oxygen ions and not only by the formation of the low occupation region. The simulated RRAM switching hysteresis cycle is in agreement with the experimental result. The main mechanism leading to the devices degradation and switching failure was identified.

(2): By varying the thicknesses of the reference in-plane ferromagnetic layers and/or the separation between them, one can modulate the switching time and achieve an almost symmetric switching in asymmetric MTJs without an external magnetic field. The simulations also highlight the importance of the field acting perpendicular to the plane. This field facilitates switching. The proposed method is used for performance optimization of STT-MRAM devices based on a dual MTJ with two barriers and ultra-thin dual MTJ cells.

(3): A new concept of a composite free layer (C-MTJ) was introduced. The composite magnetic layer consists of two half-ellipses separated by a non-magnetic spacer. As in p-MTJs, in MTJs with a composite free layer the switching barrier energy is practically equal to the thermal stability barrier. Due to the removal of the central region in the monolithic structure, the shape anisotropy is slightly decreased together with the thermal stability factor. To boost the thermal stability factor in composite structures it is sufficient to increase the thickness of the free layer and/or the aspect ratio, so the thermal stability factor exceeds that for p-MTJs demonstrated to date. An almost threefold decrease of the switching time in such structures has been found by simulation. Also a very narrow distribution of switching times is found for the proposed composite structure. In order to explain the distribution narrowing, the principles of the self-stabilization and self-acceleration of an MTJ with composite free layer switching were identified.

Furthermore, a new C2-MTJ structure with a composite free layer consisting of two ellipses with the axes a∕2 and b inscribed into a rectangle a × b was proposed and analyzed. The simulations show that, while preserving all the advantages of the C-MTJ structure, such as fast switching, high thermal stability factor, and very narrow distribution of switching times, the newly proposed structure can be easier fabricated and offers a higher potential for STT-MRAM performance optimization. A very narrow distribution of switching times of C-MTJ and C2-MTJ structures is useful not only for application in a STT-MRAM memory cell, but also for magnetic sensors.

(4): Extensive simulations show that under certain conditions there is a non-negligible probability of the formation of a vortex state/transverse domain wall during switching. This results in a drastic increase in the switching time or can even lead to a complete switching failure. The detected switching failure mechanism must be taken into account in a realistic performance optimization of STT-MRAM devices.

In addition, an exception in the criterion typically used to describe the 100% switching was found. In the case of in-plane switching from 1 to -1 the switching probability is defined equal to 1 if the normalized average magnetization M_x∕M_S along the long axis becomes less than -0.5 (M_x∕M_S < -0.5). If, however, the vortex state is generated, 100% switching can be achieved even when M_x∕M_S > 0.5. This exception demonstrates that one has to consider not only the average magnetization but also the state of the system during switching in taking decision about switching reliably.

(5): Methods for utilizing this parasitic switching effect to construct a spin-torque oscillator are shown. A newly proposed structure of the bias-field-free spin-torque oscillator has a form of a penta-layer MTJ structure with half-ellipsis reference layers. The structure with half-ellipsis reference layers develops stable oscillations with nearly constant amplitude. The Fourier transform of the signal is sharply peaked around a frequency of ~6.785GHz. It is to note that the frequency of the oscillations only slightly depends on the current value.

Furthermore, a new concept of bias-field-free spin-torque oscillators based on two MTJs with a shared free layer, which show stable oscillations without any external magnetic field has been proposed. The operating frequency of the stable oscillations can be tuned in a wide range by varying the geometry of the MTJs and the current densities flowing through the MTJs, making this structure attractive for many high frequency applications.