Maha Khademi

Research project

Voltage controlled operation and mutual synchronization of MTJ-based spin Hall nano-oscillator chains and arrays for neuromorphic computing

Project supervisor

Prof. Johan Åkerman

Recruitment date
01/09/2021

Maha Khademi

My name is Maha, I’m 25 years old and I was born in Tehran, Iran.

I have always consideredPhysics as one of the most powerful means to make the world a better place to live, So I decided to get my Bachelor’s degree in atomic and molecular Physics at Iran University of Science and Technology (IUST). Subsequently, I entered Shahid Beheshti University (SBU) to start my Master’s program in condensed matter physics, the field which is completely relevant to today’s fascinating technologies. As my master’s thesis, I worked on the Spin Hall Nano oscillator simulations and their applicationsin neuromorphic computing. I was looking forward to implementing my simulations in real life, in order to let this technology enter the industry. Then I saw the SPEAR project targeting this very goal, so I decided to join SPEAR to carry out my PhD. I think this opportunity would put me in the right place, and would help me to play my role as a researcher in physics, because by doing our own share, I believe we can make a better future. I’ve never left my hometown to live somewhere else, so I’m really excited about moving to Sweden and starting a new independent life. I really appreciate the great deal of experience that is waiting.


Project Description

The project will study nano-constriction SHNOs based on W/CoFeB/MgO stacks with and without voltage control and with and without magnetic tunnel junctions (MTJs). NanOsc has recently demonstrated mutual synchronization in both chains and two-dimensional arrays of Pt/NiFe-based SHNOs as well as voltage-controlled operation of W/CoFeB/MgO-based SHNOs. In this project, the candidate will investigate mutual synchronization in long chains and large arrays of such W/CoFeB/MgO-based SHNOs and, in particular, study whether voltage control can be used to control their mutually synchronized state. The possibilities for carrying out neuromorphic computing using such voltage controlled SHNO arrays will be explored. Their functionality with respect to thickness, composition, and perpendicular magnetic anisotropy of the CoFeB layer as well as nano-constriction width, shape, and separation will be mapped out. Non-volatile memristive control of SHNOs through the use of charge trapping inside the dielectric gate controlling the SHNO will also be attempted. Such an approach will enable the storage of local weights that can be used to train the oscillator array to match certain input patterns. The candidate will work with a large international group of excellent graduate students, postdocs, and more senior scientists, as well as the NanOsc engineers in charge of transferring these results to commercial applications.

Host institution

NanOsc is a Swedish R&D SME, developing Spin torque nano-oscillators (STNOs) and spin Hall nano-oscillators (SHNOs) for use at very high frequencies (10-100 GHz), primarily within the fields of space communication, high speed radio links, vehicle radar applications, intra/inter-chip communication, and neuromorphic computing. NanOsc’s expertise comprises the fabrication, characterization, optimization and reliability of the physical STNO and SHNO core devices. Since its foundation in 2006, NanOsc’s business is to develop intellectual property on STNOs and SHNOs, design instruments for their characterization, and with partners develop novel applications on these technologies. NanOsc benefits from close ties with the University of Gothenburg where it is one of the major external users of the MC2 cleanroom.

The project will be carried out in the laboratory of Prof. Johan Åkerman at University of Gothenburg.

Planned Secondments

CEA (Grenoble, France), under the supervision of Jean-Philippe Attané.

IMEC (Leuven, Belgium), under the supervision of Sebastien Couet.

Registering University

UGOT (Gothenburg, Sweden).

Diseño y desarrollo web Triplevdoble