VR Fundamental Fluctuations in Spintronics Final Phase
Title: VR Fundamental Fluctuations in Spintronics Final Phase
SNIC Project: SNIC 2022/5-425
Project Type: SNIC Medium Compute
Principal Investigator: Gunnar Malm <gunta@kth.se>
Affiliation: Kungliga Tekniska högskolan
Duration: 2022-09-29 – 2023-02-01
Classification: 21001
Homepage: https://www.kth.se/profile/gunta/page/noise-and-fluctuations


The purpose of this project proposal is to study fundamental fluctuations in microwave spintronic devices, caused by process variability. The research is supported by the GRANTED VR project 2018-21, PI Assoc. Prof. Gunnar Malm, KTH 2017-04196 - Fundamentala Fluktuationer i Spinntronik A PhD student has been active in the project since 2018 and is planned to graduate in spring 2023. Several manuscripts under preparation fall 2022 will require HPC resources. These fluctuations in spintronics will be investigated in terms of e.g. frequency stability, oscillator linewidth, and mode jumping. A crucial part of the project is to achieve improved STO reliability and fidelity of the simulation models that are used to predict and analyze the experimentally observed characteristics. The active part of a spintronic device is an ultra-thin ferromagnetic layer stack. The inferior or varying quality of this layer stack, which is due to process variability, is the root cause of device fluctuations. The layer physical quality can be analyzed by various complementary methods. In our project such physical data, from in-house fabricated sample devices, will be fed into micro-magnetic simulations in an eScience approach, where massive simulations are performed on dedicated computer hardware using tailored software and distributed computing on clusters. In our specific design-of-experiment (DOE) we will input individual randomized devices, with realistic properties to the simulator, and look at the variability in the output. The simulations are designed to complement our experimental work on spin torque oscillators and increase the understanding of the impact of fabrication defects device-to-device variability and noise processes. In order to achieve this, we need to perform a massive amount of simulations that will heavily rely on the quality of the measured data. Since a single simulation takes from hours to days and a sufficient sample size is crucial for meaningful statistical data extraction, it is of utmost importance to work with highly efficient and parallelizable software.