All-Optical Characterisation and Wavefront Manipulation of Photonic Integrated Circuits

Abstract

Advanced photonic probing techniques are of great importance for the development of non-contact wafer-scale testing of Photonic Integrated Circuits (PICs). The most advanced characterisation of a PIC is given by the electromagnetic field distribution of the light propagating within it. Obtaining such a complete description typically requires complex photonic probing techniques which utilise a scanning perturbation placed in the near-field to infer information on the fields within. However, such approaches are limited with most real-world devices due to high-cost, low-throughput testing and the use of protective cladding layers that block access to the near-field. Here, we demonstrate Ultrafast Photomodulation Spectroscopy (UPMS) as a far-field technique for the non-destructive characterisation of individual PIC elements. A scanning optical pump creates a highly localised perturbation in the refractive index profile of a silicon waveguide via free-carrier excitation, which can be raster-scanned over the device to map the internal electric field distributions without requiring direct access to the near-field. Sub-micrometer resolution is obtained, and direct comparison to a rapid and rigorous analytical model enables the quantitative comparison to an ideal design structure. In addition to constituting a promising route for the testing and diagnostics of PICs, the pump-probe system can also be used for the real-time trimming of photonic devices with active feedback. The optical annealing of ion-implanted racetrack resonators is shown here to be capable of locating the critical coupling position with an accuracy of ±200 nm. Multiple refractive index perturbations can be used for intricate wavefront manipulation of photonic structures with complex functionalities, such as optical routers, modulators, (de)multiplexers and mode converters. In this work, perturbation patterns are initially designed using brute-force iterative techniques, leading to the development of a deep-Artificial Neural Network (ANN) that is capable of rapidly ( 1 ms) and accurately (<6% mean transmittance error) generating a new perturbation patterns with arbitrary complex transmission matrices for universal optical components.

More information

Identifiers

ORCID for Nicholas Dinsdale: orcid.org/0000-0001-9870-5700

Catalogue record

Export record