Deep Learning-Trained Imager Magnifies Subwavelength Objects

An optical imaging system from UCLA goes beyond the traditional diffraction limit to enable imaging at subwavelength resolution. The new imager will make direct imaging of phase objects with subwavelength resolution less challenging for bioimaging, sensing, material characterization, and other applications that frequently use phase imaging.

The imager, developed in the lab of UCLA professor Aydogan Ozcan, enables subwavelength imaging of phase and amplitude objects. To enable the imager to recover high-frequency information corresponding to the subwavelength features of an object, the research team uses all-optical diffractive encoding and decoding with a solid-immersion layer.

The imager’s thin, high-index, solid-immersion layer transmits high-frequency information about the object to a spatially-optimized diffractive encoder. The encoder converts and encodes the high-frequency information into low-frequency spatial modes for transmission through air.

A diffractive decoder, which is jointly trained with the encoder surface, processes the encoded spatial information that is propagated through the air to create a magnified image of the input object. The magnified image reveals subwavelength features that would normally be washed out due to diffraction limitations.

A team of researchers, including professor Aydogan Ozcan at UCLA, developed a new method for achieving subwavelength-resolution imaging for phase and amplitude objects. The technique relies on diffractive encoding and decoding with a solid-immersion layer to recover high-frequency information corresponding to the subwavelength features of an object. Courtesy of UCLA.

To demonstrate the subwavelength diffractive imager, the researchers fabricated a multilayer, monolithic design that operates at the terahertz (THz) part of the spectrum. They tested this monolithic diffractive encoder-decoder pair with a customized, high-resolution THz imaging system. The experimental results confirmed that the 3D-fabricated, solid-immersion diffractive imager can resolve phase objects by directly performing transformations through the diffractive encoder-decoder pair.

At THz frequencies, the imager can resolve features as small as λ/3.4 (where λ is the illumination wavelength) by directly transforming them into magnified features at the output.

The trained subwavelength diffractive imager generalized to previously unseen objects from the same distribution as the objects used in training, demonstrating internal generalization. It also generalized to new types of objects from completely different datasets, demonstrating external generalization capability.

The user can operate the subwavelength imager at different parts of the electromagnetic spectrum by physically scaling — that is, by expanding or shrinking — the optimized diffractive features of the encoder and decoder surfaces in proportion to the illumination wavelength. This can be done without needing to redesign the diffractive features of the system.

The subwavelength imager offers the advantage of directly performing quantitative phase retrieval, eliminating the need for lengthy computer processing, which consumes a lot of power.

The researchers believe that the solid-immersion diffractive imager, with its compact size, cost-effectiveness, and ability to capture subwavelength features, could lead to significant advancements in bioimaging, sensing, and material inspection, among many other applications.

The research was published in eLight (www.doi.org/10.1186/s43593-024-00067-5).