Plasmonic ommatidia for lensless compound-eye vision

Traditional cameras, while offering high spatial resolution, face a tradeoff between size and field of view due to aberration effects. Nature's solution, the compound eye found in insects and crustaceans, overcomes this limitation through an array of ommatidia, each capturing a single point of light information from a different direction. This architecture allows for compact size, wide-angle vision, and high motion sensitivity, making it ideal for applications such as endoscopy, surveillance, wearable cameras, and autonomous navigation. Previous attempts at replicating compound eyes using microlenses faced challenges due to the curved geometry and compatibility issues with planar microelectronics. Flat compound eyes have limited field-of-view due to optical crosstalk. This research presents a novel approach using a lensless planar architecture which eliminates these limitations.

Existing artificial compound eye cameras primarily use planar or curved arrays of microlenses coupled with image sensors. Curved geometries, while closely mimicking natural compound eyes, are difficult to integrate with standard planar microelectronics, requiring bulky optical relays or complex fabrication processes. Planar designs suffer from limited field-of-view due to inter-pixel crosstalk, even with multiple lens arrays. Other approaches utilize diffraction gratings or optical phased arrays for angle-sensitive detection, but these methods have limitations in field-of-view or require complex deconvolution or laser oscillators. A recent design used nanowires for angle sensing, but was limited to triangulation and not full image reconstruction. This work proposes a fundamentally different approach that addresses these limitations.

The proposed compound-eye camera uses a planar array of pixels, each coated with an angle-sensitive metasurface. This metasurface, composed of metallic plasmonic nanostructures, transmits light only within a small, geometrically tunable angular range. The design utilizes three key sections: a periodic grating coupler, a grating reflector, and slits in the underlying metal film. Incident light at the desired angle is diffracted into surface plasmon polaritons (SPPs), which are then scattered into the photoconductive Ge substrate, generating a photocurrent. The angle of peak detection is controlled by the grating coupler period. Gold is used for its favorable plasmonic properties in the near-infrared. Full-wave electromagnetic simulations using the finite-difference time-domain (FDTD) method were employed to optimize the metasurface design parameters, such as grating coupler period, number of nanoparticles, and nanoparticle width. Experimental devices were fabricated using metal-semiconductor-metal (MSM) Ge photoconductors. Angle-resolved photocurrent measurements were performed using a custom-built optical goniometer setup to validate the design, measuring the photocurrent response as a function of both polar and azimuthal angles of incidence. Image reconstruction was performed using a linear matrix equation (y=Ax) where the object's intensity distribution (x) is related to the captured data (y) through a sensing matrix (A) containing the angular responses of all pixels. The truncated singular value decomposition (TSVD) technique was used for image reconstruction. Simulations were conducted with both calculated and measured angular response patterns, incorporating Gaussian noise to account for realistic sensor performance. The impact of broadband illumination was assessed by incorporating a Gaussian blurring kernel to model the broadening of the detection peak due to wavelength variations.

FDTD simulations showed that the designed metasurfaces achieve tunable directional photodetection with a wide tuning range (±75°) and narrow angular resolution (3-14° FWHM). The peak transmission coefficient was in the range of 35-45%, with a peak-to-average-background ratio of about 6 for p-polarized light. Measurements of fabricated devices demonstrated highly directional response in good agreement with simulations, showing C-shaped regions of high responsivity. The measured polar-angle selectivity was in the range of 4-21° FWHM. The peak p-polarized responsivity was reduced to ~42% and 36% of that from a bare sample at peak detection angles of 12° and 65°, respectively. Image reconstruction simulations, using both simulated and measured angular response data, demonstrated the ability to reconstruct high-quality images of relatively complex objects over a wide field-of-view (±75° for simulated, ±65° for measured) with realistic signal-to-noise ratios (SNR) of 56, 63, and 73 dB. Simulations accounting for broadband illumination (δλ/λ₀ = 5% and 10%) also showed successful image reconstruction with only a minor loss of resolution.

The results demonstrate the successful development and characterization of a novel family of angle-sensitive photodetectors that mimic the functionality of apposition compound eye ommatidia in a planar, lensless format. This technology offers the advantages of small size, wide field-of-view, and high temporal bandwidth. The achieved field-of-view exceeds those of previous planar implementations and is comparable to curved designs while significantly simplifying fabrication. The use of Ge MSM photoconductors was convenient for prototyping, but the design is compatible with other materials and photodetector technologies, including Si photodiodes. Future work could focus on using dielectric metasurfaces for reduced losses and CMOS compatibility, creating polarization-independent devices and broader bandwidth operation. Advanced computational imaging techniques, such as iterative reconstruction algorithms or machine learning, could further improve image quality and capabilities. This research highlights the potential of combining optical metasurfaces and computational imaging for innovative applications.

This work presents a novel lensless compound-eye camera architecture based on plasmonic metasurfaces integrated with standard image sensor arrays. The research successfully demonstrated the ability to reconstruct high-quality images over a wide field of view with realistic noise levels. This platform is adaptable to various materials and technologies, paving the way for compact, high-performance imaging systems with applications across numerous fields. Future research could explore more advanced metasurface designs for enhanced performance, broader bandwidth operation and further integration with computational imaging algorithms.

The current experimental devices show some discrepancies compared to simulations, mainly due to surface roughness and variations in fabrication parameters. The polarization dependence of the devices may limit sensitivity for unpolarized light, although it could be exploited for polarization vision. The maximum field-of-view in the experiments was limited by the fabricated devices, but could be further extended. The use of a relatively simple image reconstruction algorithm could be improved upon using more advanced techniques.