Energy-efficient non-volatile memory: Resonant SAW assisted STTRAM
Principal Investigator: Jayasimha Atulasimha, Qimonda Professor, Department of Mechanical and Nuclear Engineering
Institution: Virginia Commonwealth University
Energy Efficient Non-Volatile Random Access Memory: Resonant SAW Assisted STTRAM
1. Introduction: We propose an energy efficient non-volatile random access memory (RAM) device that will enable at least an order of magnitude lower energy dissipation compared to the conventional spin transfer torque (STT) RAM while switching at ~1ns (~1GHz).
(a) Our Concept and existing/competing approach: Magnetic tunnel junction (MTJ) based on a ferromagnet/oxide barrier/ferromagnet structure (for example, CoFeB/MgO/CoFeB) is an existing commercial magnetic memory technology called STT-RAM. In such devices (Fig 1 a), a current passing through a fixed layer is spin polarized and switches the magnetization of the free layer to write a bit of information. Unfortunately, the energy dissipated in such switching is ~100 fJ/bit , which is 100 to 1000 times higher than the energy dissipated to switch a CMOS device. This is an enormous cost to pay for non-volatility! The key idea of this proposal is to use resonant surface acoustic wave (r-SAW) in conjunction with STT-current to significantly decrease the STT current needed to write a bit and consequently increase the energy efficiency of these magnetic memory devices. We claim and show that when resonant SAW is applied to the soft layer of a perpendicular MTJ (p-MTJ) array shown in Fig 1, it creates a cyclic stress that forces the magnetization to precess around the perpendicular magnetic anisotropy (PMA) direction with large deflection from it. This is essentially acoustically induced ferromagnetic resonance. The higher this deflection, the lower the STT current required to switch the p-MTJ.
Figure 1: (a) Magnetic tunnel junction (Fig. 1 (a) is reproduced from Ref ). (b) MTJ array deposited on a piezoelectric substrate.
(b) Significant features and novelty of our approach: The promise of this invention is it only involves a minor modification to existing memory technology but can lead to at least about 10 times reduction in energy consumed, which would make it viable for industry to adopt. This minor modification could be realized by introducing interdigital transducer (IDT) fingers and the piezoelectric substrate over which the p-MTJ arrays are fabricated. Application of suitable electric pulses on the IDT electrodes can produce a SAW that can strain the soft layer of the p- MTJs dynamically and thus reduce the STT current requited to switch it.
2. Working Principle of Proposed Idea: Initially, the magnetization points in the out-of-plane direction (+z-axis) in Fig. 2, and a SAW train (few cycles) was applied across the magnets to find the resonant frequency, which is dependent on the anisotropy of the nanomagnets and magnitude of the SAW (as the oscillation is non-linear). This SAW applies compression and tension along the y-axis of the nanomagnet (and vice versa along the x-axis). This cyclical stressing of the magnet rotates the magnetization from pointing directly out-of-plane to begin to precess about the z-axis, as shown in Fig. 2. As a result of the resonant SAW, the cone of rotation becomes larger with time and the magnetization approaches the x-y plane of the nanomagnet. Spin-transfer-torque was applied after the maximum deflection of the magnetization from the out-of-plane direction was achieved. Once the magnetization is switched and the STT current is withdrawn, the SAW can be concurrently removed or continued to run for a few cycles before withdrawal, as it is not enough to switch the magnetization on its own. NOTE: The energy to generate the SAW is amortized over many nanomagnets and is hence negligible. More importantly, the industry challenge is to reduce the write current that will allow smaller write transistors (see Fig 1 a).
Figure 2: Magnetization dynamics with resonant SAW + STT switching of out-of-plane magnetization.
3. Preliminary results: Our preliminary simulations  under a macro-spin assumption (whole nanomagnet acts like a single classical magnetization pointing in one direction) reports that ~ 99.99% switching is attainable for p-MTJs when SAW and STT are applied in combination with ~10 times lower energy dissipation compared to conventional STT-current approach.
4. Technical Approach: We will carry out rigorous micromagnetic simulations and perform key experiments to investigate the feasibility and realistic challenges of our proposed approach.
4.1 Micromagnetic simulation: We will build on our preliminary macrospin (the entire nanomagnet is treated as a giant classical magnetization) simulations to model the device in micromagnetic framework (allows the magnetization of the nanomagnet to break into several regions of tiny magnetization that are not parallel to each other) and understand the effect of realistic material inhomogeneity such as grains, surface roughness and edge effects on magnetization dynamics using micromagnetic simulation (Mumax). Here, the soft layer of the p-MTJs will be simulated using the Landau-Lifshitz-Gilbert (LLG) equation in the presence of thermal noise at room temperature. The effective magnetic field (Heff) driving the magnetization dynamics is given by:
Here, Hdemag is the effective field due to demagnetization energy, Hexchange is the effective field due to exchange coupling. Strain effectively modulates the anisotropy energy and is incorporated by modulating Hanisotropy. Thermal noise will be modeled by a random, effective magnetic field (Hthermal) applied in the micromagnetic framework. When strain is used in conjunction with STT, the STT effects will be modeled with the appropriate terms. This modeling framework will be used to evaluate the optimal geometry and material system as well as investigate incoherent magnetization and its impact on the switching of magnetization.
4.2 Experiments to realize SAW+STT memory device: We will fabricate and pattern nanodots with perpendicular anisotropy along with the interdigitated electrodes to launch the SAW. At VCU using the Virginia Microelectronics Center (VMC) and Nanomaterials Core Characterization (NCC) we will deposit the film and perform e-beam lithography. We will pattern this film into nanoscale devices and electrodes to launch the SAW at NIST, Gaithersburg (where we have academic user rates). We will test these memory elements with applying a combination of perpendicular magnetic fields (instead of STT in this preliminary demonstration) in combination with SAW. We will use magnetic force microscopy (MFM) at VCU to show that the SAW reduced the magnetic field needed to switch the proposed memory device.
5. Expected Results: This work will advance the development of energy-efficient magnetic memory devices by putting our idea on a firm footing.
6. VMEC Industry Interest, Collaboration and Impact on Virginia: In addition to the new knowledge and physics generated by this research effort, the results could lead to the development of very low power magnetic memory devices. The PI is willing to partner and collaborate with Micron (part of VMEC consortium) if there is mutual interest on pursuing this technology. Preliminary discussions with Micron managers (at Boise) indicated they found this idea of interest. In preparing larger grants to DOD, the PI could partner with Virginia universities (e.g. UVA) where there is a mutual interest.
7. Approximate Cost: The cost includes support (stipend, tuition will be paid from another source) for 1 PhD student for 4 months ($9 K), 1 week (0.25 month) PI summer salary ($4.5 K), clean room/equipment cost ($1K) and modest travel to NIST ($0.5 K). This would provide preliminary data to write large grants to NSF and DOD to further develop this idea. The PI has no funded grants on this topic.
 Wang, K. L., Alzate, J. L., Khalili Amiri, L., (2013). Low-power non-volatile spintronic memory: STT-RAM and beyond, Journal of Physics D: Applied Physics, 46, 074003.
 Roe, A., Bhattacharya, D., Atulasimha, J. (2019). Resonant acoustic wave assisted spin- transfer-torque switching of nanomagnets. Applied Physics Letters, 115(11), 112405.