2019 Fiscal Year Research-status Report
Dynamics-based neuromorphic properties of spintronics devices
| Project/Area Number |
19K15428
|
|---|
| Research Institution | Tohoku University |
|---|
Principal Investigator |
KURENKOV ALEKSANDR 東北大学, 先端スピントロニクス研究開発センター, 学術研究員 (80830645)
|
|---|
| Project Period (FY) |
2019-04-01 – 2021-03-31
|
| Keywords | Spintronics / Spin-orbit torque / Fast switching / Antiferromagnet / Neuromorphic / Synapse / Neuron |
|---|
| Outline of Annual Research Achievements |
In FY2019 I and collaborators have advanced in the understanding of the switching in antiferromagnet/ferromagnet heterostuctures. The switching was found to be governed not only by the symmetries of spin-orbit torque, Dzyaloshinski-Moriya interaction and magnetic field but also exchange bias. It varies throughout the sample creating complicated energy landscape which promotes domain wall-driven magnetization reversal in some directions and suppresses it in the others. As a result, different areas of the ferromagnet switch either deterministically, stochastically or remain fixed. Control of the stack properties such as exchange bias allows achieving the desired switching properties which opens the way towards ultrafast neuromorphic devices including synapses and neurons.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
The ultrafast neuron and synapse proposed in the application rely on magnetization dynamics. Thus understanding it in details is crucial for the project. Initial simulations assumed coherent rotation of magnetization in each binary domain of the memristive anftiferromagnet/ferromagnet structure. However the experiment and following analysis have revealed that switching is dominated by domain expansion in the preferable directions determined by exchange bias landscape. Therefore, the model needs to be reconsidered to account for the non-coherent switching.
Influence of the ferromagnetic and antiferromagnetic layers properties on memrisrtivity, areal "completeness" of switching and switching current have been determined. It will allow optimizing the stacks for artificial neurons and synapses.
|
| Strategy for Future Research Activity |
Next steps include micromagnetic simulation of the magnetization response to very short pulses (~100 ps) in the antiferromagnet/ferromagnet structures. It is necessary to clarify how the domain pattern will respond to several pulses applied in sequences, characteristic for synapses and neurons. Despite the non-coherent magnetization reversal, short-term memory properties required for synapses and neurons are expected. Moreover, this switching mechanism may provide more biologically plausible results than in the case of independent switching of different areas of the sample. Micromagnetic simulations will be performed in mumax3 software package. This step will clarify optimal parameters of the electric pulses and allow experimental proceeding with samples emulating synapses and neurons.
|
| Causes of Carryover |
As described in the project progress section, it has been experimentally discovered that adjustments need to be done in the model describing our material system. Therefore the grant funds were invested in the hardware for micromagnetic simulations required for deciding the exact experimental parameters. Purchase of the experimental equipment has been postponed until the simulations are complete. This is the reason for carrying over the funds to the next year. The planned usage is on the experimental equipment with properties deduced from the ongoing simulations.
|
