< Terug naar vorige pagina

Project

Ferro-elektrische akoestisch-optische synaptische technologie (FAST) (FAST)

Efforts are being made to develop bio-inspired optical neuro-synaptic networks to support “neuromorphic computing” for ultra-fast communication. The idea is to split the information of an optical signal using an “acousto-optic modulator” (synapse) into multiple signals. The signals are then communicated through multiple parallel channels similar to neurons and recombined at the receiver end by using another acousto-optic modulator. Such neuro-synaptic networks are targeted for less than 100 mW power budget and greater than 100 GHz communication bandwidths which are currently hindered by the performance of the piezoelectric transducer used in the acousto-optic modulators. It generates high-frequency acoustic waves inside the attached waveguide which act as dynamic optical gratings offering periodic modification of the refractive index of the medium through which the light is traveling. To solve this issue, the applicant proposes exploring optically birefringent ferroelectrics with strong elasto-optic coefficients as stand-alone acousto-optic modulators. The project aims for the development of a CMOS compatible 100 GHz-THz range ferroelectric acousto-optic modulator (FAOM) and it's integration in a self-learning neuro-synaptic network. The applicant will unite his expertise in materials engineering and photo-ferroelectrics with Katholieke Universiteit Leuven’s (KUL) and IBM’s knowledge of the integration and characterization of neuromorphic networks with the following objectives:
1. Growth of CMOS compatible strain-engineered BaTiO3 (BTO) heterostructures.
2. Fabrication and characterization of two CMOS integrated ferroelectric (BTO) acousto-optic modulator designs based on (2.1) photo-induced acoustic strain gradients by optical control of ferroelectric domains and (2.2) acoustic confinement.
3. Demonstration of FAOM-based self-learning optical convolution neuro-synaptic network.

Datum:1 aug 2020 →  31 jul 2022
Trefwoorden:Photonic neurosynaptic networks, acousto-optic modulators, self-learning, ferroelectric domains
Disciplines:Neuromorphic computing