2020-932 Terahertz Pulse Shaping Using Diffractive Legos

SUMMARY

UCLA researchers in the Department of Electrical and Computer Engineering have developed an all-optical diffractive network. This learning-based diffractive pulse engineering framework utilizes deep learning and wave-optics to design an arbitrarily shaped broadband pulse into a desired waveform across a broad bandwidth and high spectral resolution.

BACKGROUND

Machine learning is being applied to optics to make progress in optical pulse shaping for a variety of data processing problems. Pulse shaping technologies have limited applicability at the terahertz band spectrum due to a lack of innovation within spatio-temporal modulation and complex wavefront controls space. As a result, terahertz pulse manipulation has been performed indirectly through the engineering of optical-to-terahertz converters or shaping of the optical pulses that pump these terahertz source. New methods are needed that allow for the modulation and control of complex spatio-temporal wavefronts, while providing high spectral resolution and a broad bandwidth at these frequencies.

INNOVATION

UCLA researchers from the Department of Electrical and Computer Engineering have developed a modular pulse shaping network that uses machine learning to directly modulate/modify terahertz pulse frequencies. Direct manipulation of terahertz pulses has been demonstrated here using diffractive networks that can shape various temporal waveforms of interest. The method can modify a terahertz pulse independent of polarization, beam shape or quality, and aberrations. The method optically shapes pulses by simultaneously controlling relative phase and amplitude of spectral components using trainable diffractive layers. This allows for a small chip footprint footprint and compact pulse engineering system.

POTENTIAL APPLICATIONS

• Design an arbitrarily shaped broadband pulse into a desired waveform across a broad bandwidth and high spectral resolution.
• Tele-communication
• Ultrafast Imaging
• Spectroscopy
• Pulse Compression
• Wave fields in the 4D spatio-temporal space
• All-optical machine learning platform

ADVANTAGES

• Ability to design new wave forms within limited time frame
• Direct application to generate terahertz pulses whilst, using with other fabrication methods and materials
• Versatility, modular, and easy adaptability to engineer terahertz pulses irrespective of their polarization state, beam quality as well as spectral/spatial aberrations

RELATED MATERIALS

• Veli, Muhammed, et al. “Terahertz Pulse Shaping Using Diffractive Surfaces.” ArXiv.org, 21 Nov. 2020, arxiv.org/abs/2006.16599.

STATUS OF DEVELOPMENT

This learning-based diffractive pulse engineering framework has been tested and validated.