Graphene plasmonic spatial light modulator for reconfigurable diffractive optical neural networks

Area of Science:

• Optics and Photonics
• Materials Science
• Artificial Intelligence

Background:

• Terahertz (THz) diffractive optical neural networks (DONNs) offer simultaneous image capture and classification.
• Current DONNs primarily use passive components, limiting their functionality.
• Reconfigurability via spatial light modulators (SLMs) is desired, but device requirements and implementation challenges are unclear.

Purpose of the Study:

• To investigate the use of an active reflective graphene-plasmonics-based SLM for reconfigurable THz DONNs.
• To determine the necessary device specifications for achieving complex-valued modulation in THz DONNs.
• To demonstrate the feasibility and performance of such reconfigurable THz DONNs.

Main Methods:

• Developed an active reflective SLM utilizing graphene nanoribbons and a back reflector.
• Coupled plasmonic resonance with Fabry-Pérot (F-P) modes to achieve specific amplitude and phase modulation (π/2).
• Constructed and tested reconfigurable reflective THz DONNs with the designed SLMs on the MNIST dataset.

Main Results:

• Achieved a substantial π/2 phase modulation and minor amplitude modulation using the graphene-plasmonics SLM.
• The developed reconfigurable THz DONNs demonstrated over 94.0% validation accuracy on the MNIST dataset.
• Confirmed the effectiveness of complex-valued modulation for DONN reconfigurability.

Conclusions:

• Active reflective graphene-plasmonics SLMs enable effective reconfigurability in THz DONNs.
• Relaxed SLM specification requirements can simplify design and enable diverse DONN functionalities.
• This approach advances intelligent THz imaging and classification capabilities.