Computing with nonlinear spin-wave dynamics
|Contact: DE-LOUBENS Gregoire, , firstname.lastname@example.org, +33 1 69 08 71 60|
In magnetic nanostructures, the excitation eigenmodes (spin-waves) are coupled together via nonlinear interactions. The main idea is to use this dynamical system to perform neuromorphic computing tasks.
|Possibility of continuation in PhD: Oui|
|Deadline for application:08/04/2021 |
|Full description: |
Spin-waves (SWs) are the collective excitations of magnetization in ferromagnets. Their natural frequency is typically in the GHz range with propagation lengths over several microns depending on the intrinsic damping of the material. Due to exchange and dipole-dipole interactions, their dynamics is inherently nonlinear and can exhibit rich physics. In confined geometries like thin film waveguides and dots, SW modes are quantised with frequency spacings controlled by the lateral dimensions of the magnetic sample, which can be further modified by external stimuli such as applied magnetic fields or spin transfer torques. Large amplitude stimuli can trigger nonlinear processes like mode conversion and mode instabilities, resulting in the redistribution of energy between coupled SW modes [1,2].
During this internship, we will investigate experimentally the capacity of SWs in nanostructured thin films to perform reservoir computing . The basic control mechanism is the nonlinear coupling between SWs, which allows orthogonal eigenmodes of the equilibrium state to interact with each other as their amplitudes increase. Because such coupling involves thresholding events , like for spiking neurons, we can achieve computational tasks with a cognitive nature like classification. For this, we will perform a multifrequency spectroscopy of ultra-low damping magnetic nanostructures in the nonlinear regime . We will use a magnetic resonance force microscope, a home made near field technique able to sensitively detect SW dynamics in individual nanomagnets . To analyze the experimental results and identify configurations useful for reservoir computing based on recurrent neural network, we will also rely on micromagnetic simulations based on an open source python code . In the mid-term, this might allow for a new hardware implementation of reservoir computing that relies on the liquid state machine concept  at GHz frequencies, which could be useful for processing telecommunications signals.
This internship will take place in the context of two recently funded projects, one by Europe (k-NET), and another one by the French ANR (MARIN), and will therefore be conducted in a collaborative environment.
 V. Naletov et al., Ferromagnetic resonance spectroscopy of parametric magnons excited by a four-wave process, Phys. Rev. B 75, 140405 (2007)
 Y. Li et al., Nutation Spectroscopy of a Nanomagnet Driven into Deeply Nonlinear Ferromagnetic Resonance, Phys. Rev. X 9, 041036 (2019)
 W. Maass et al., Real-time computing without stable states: A new framework for neural computation based on perturbations, Neural Computation 14, 2531 (2002)
 O. Klein et al., Ferromagnetic resonance force spectroscopy of individual submicron-size samples, Phys. Rev. B 78, 144410 (2008)
 C. Fernando & S. Sojakka, Pattern Recognition in a Bucket in Lecture Notes in Computer Science, vol 2801 (2003)
|Technics/methods used during the internship: |
Magnetic force microscopy; high frequency techniques; micromagnetic simulations
|Tutor of the internship |