Open positions between 2024

PhD thesis: Artificial neural network constructed using multiple coupled phononic modes for per-reservoir computing

In recent years, neural-inspired computing, such as reservoir computing, has become an interesting topic for understanding how an artificial neural network works and the contribution of dynamic memory to the computational function. These open questions will lead to the development of the next generations of AI hardware. So far, reservoir computing has been achieved by constructing a network of time-multiplexed virtual nodes sharing a single nanomechanical resonator, with classical Duffing nonlinearity as the source of nonlinearity [1].

In this thesis, we propose to extend single nanomechanical resonators to parametrically coupled phonon modes in order to perform reservoir calculations in multimode systems. This new configuration will significantly improve the computational capacity of the artificial network, instead of using the current technology of visual modes. The objectives of this work include (1) the use of the COMSOL multiphysics simulator to design phononic crystal structures on a suspended membrane resonator in order to achieve the quality factor and resonant frequency parameters of the mechanical and phononic modes [2]. (2) Simulations of an artificial neural network constructed by parametrically coupled multiple modes composed of mechanical and phononic modes. (3) fabrication of samples in a clean room, taking advantage of the process that has been well developed by the group. (4) High-frequency measurement of the reservoir calculation by calibrating the non-linearity of the parametric coupled modes, testing the benchmark in both measurement and simulation and measuring the error rate and memory capacity of the neural network. (4) High-frequency measurement of the reservoir calculation by calibrating the non-linearity of the parametric coupled modes, testing the benchmark in both measurement and simulation, and measuring the error rate and memory capacity of the neural network. The whole measurement will be performed based on a microwave optomechanical scheme which has been well developed in the team [3].

This project is highly interdisciplinary and will enable the PhD student to benefit from advanced technologies in numerical simulation, nanofabrication and microwave engineering. It will provide access to the frontiers of neural-inspired computing, phononic structures and optomechanical technologies. The student will be strongly supported by both researchers and permanent engineers from the physics group and the nanofabrication platforms.

Ref: [1] JOURNAL OF APPLIED PHYSICS 124, 152132 (2018),  [2] Phys. Rev. Appl. 14, 024068 (2020),  [3] Nano Letter, 21 (13), 5738–5744 (2021)

Requirement : basic knowledge of electromechanics, microwave engineering OR  back ground of analytical/numerical simulations

Starting date: 2024-2025

Benefits:  excellent clean-room facilities, microwave optomechanical platform, cryogenics, COMSOL simulators

If you are interested in these topics, please contact me by emal:  xin.zhou @cnrs.fr