The rapid development of information technology (IT) has had profound consequences on most aspects of our civilization. However, the energy consumption associated to IT continues to rise at an accelerated pace. In this context the microelectronics industry faces major challenges related to power dissipation and energy consumption. A promising solution is the integration of non-volatile random access memories. This minimizes static power and allows normally-off/instant-on computing.
Spin Transfer Torque Magnetic Random Access Memory (STT-MRAM) has been identified by the ITRS as the best candidate. STT-MRAM memory cell is formed by two ferromagnetic layers (“fixed” and “free” layer”) separated by an insulator – a magnetic tunnel junction (MTJ). The information is stored in the magnetization of the “free” layer. The magneto-resistance of the MTJ depends on the orientation of the free layer. For reading, a low bias voltage is applied while for writing, a larger voltage is required; The electric current, spin polarized by the fixed layer, carries spin angular momentum to the free layer (hence the name Spin Transfer Torque – STT). Inherent to its working principle, this memory element has high density, good scaling and low consumption. However, during the writing process, the voltage can damage the thin insulating layer, limiting the speed as well as the endurance.
We use an alternative way to switch the magnetization, not by spin transfer from a ferromagnet, but by transferring angular momentum directly from the crystal lattice. The Spin-Orbit Torque (SOT) occurs in materials with large spin orbit coupling that lack structural inversion symmetry. From a practical perspective, the most significant difference is related to the geometry of the current injection. Instead of passing the current vertically, like in STT, the SOT switching relies on in-plane current injection. This improves the reliability of the memory element since there is no electric stress on the tunnel barrier during the writing phase.
Based on this difference, we propose to use the liberty allowed by the in-plane current injection to design novel magnetization switching schemes. We have discovered that the lateral shape of the devices, as defined by lithography during the device fabrication, can be used to determine the SOT induced switching of the magnetization. Magnetic objects having different shape will react differently to the electric current. For example, depending on their shape, two such objects can switch in opposite directions when subjected to the same current pulse.
Magnetic images of the shape controlled switching by current pulses. (initial state – top; final state – bottom)
We have studied broadly the material dependence of the SOT. We have evidenced different physical mechanisms inducing SOT, originating either from the interface of the heavy metal or from the bulk of the thin film. we also discovered that SOT can also be created in a more subtle manner, as a consequence of the asymmetric confinement of the electronic states between two oxides, without any heavy metal layer.
From the perspective of magnetization dynamics, we have designed switching schemes where the magnetization reversal is controlled by the geometry [Safeer, C. K., et al. “Spin–orbit torque magnetization switching controlled by geometry.” Nature nanotechnology 11.2 (2016): 143-146.]. These have direct applications for the SOT-MRAM technology, as they allow zero-field switching [Patent no: FR 15 59584 (2015); FR 15 59914 (2015); FR 21 05175 (2021) ]. The lateral shape of the device breaks the same mirror symmetry as the in-plane magnetic field required for deterministic SOT switching. The underlying physical phenomenon allowing for such effects, is the special asymmetric dependence of the domain wall (DW) velocity on the angle of the injected current. We have discovered that that this asymmetry is affected by two factors: the well known Dzyaloshinskii–Moriya interaction, as well as the recently reported Chiral Damping. This last ingredient has proven to be essential not only for the DW dynamics it is also critical for the Skyrmion stability and motion.
Finally yet importantly, we have shown that the shape dependence of the DW motion can be used for performing neuromorphic computing [Patent no: EU 21306379.5 (2021)].
Contacts: Mihai MIRON