Microscale Infrared Technologies for Spectral Filtering and Wireless Neural Dust

Abstract

Pivotal technologies, such as optical computing, autonomous vehicles, and biomedical implantables, motivate microscale infrared (IR) components. Hyperspectral imagers (HSI), for example, require compact and narrowband filters to obtain high-spatial and -spectral resolution images. HSIs acquire continuous spectra at each pixel, enabling non-destructive analyses by resolving IR scattering/absorption signatures. Toward this end, dielectric subwavelength gratings (SWG) are intriguing filter candidates since they are low-loss, have no moving parts, and exhibit narrow spectral features. Wireless neural implantables are another apropos microscale IR technology. Wireless IR data and power transfer disposes of infection-prone percutaneous wires by leveraging the IR transparency window in biological tissue. This dissertation contains two related topics. The first details SWG IR filters, and the second studies progress toward wireless neural motes.

This work extends the capabilities of SWG IR filters. Following a theoretical overview, mid-wave infrared (MWIR, 3-7 um) transmittance filters are experimentally demonstrated using the zero-contrast grating scheme. Via a facile silicon fabrication process, we realize narrowband polarization-dependent and polarization-independent MWIR transmittance filters with some of the highest Q observed in MWIR SWGs. An empirical study confirms the relationship between filter performance and grating size, an important trade-off for HSIs. We then demonstrate GaAs SWG filters for monolithic integration with active optoelectronic devices. The GaAs SWGs perform comparably to their silicon counterparts.

To enable narrowband filtering at normal incidence, we investigate symmetry-breaking in geometrically asymmetric gratings. The presented SWG geometries access quasi-bound states in the continuum (BIC). Studies in Fano resonance and diffraction efficiency symmetry provide physical insight. Asymmetric 1D and 2D SWGs furnish polarization-dependent and -independent filtering, respectively. We experimentally demonstrate normal incidence long-wave IR (LWIR, 7-12 um) transmittance filtering in asymmetric SWGs and confirm symmetry-breaking implications. A reduced-symmetry hexagonal pattern presents an early design for truly polarization-independent quasi-BIC coupling in SWGs.

Advancements in implantable neural devices promise great leaps in brain mapping and therapeutic intervention. To meet this challenge, we investigated a wireless neural mote system using near-infrared (NIR, 800 nm – 3 um) photovoltaics and LEDs to wirelessly harvest power and transmit data. The neural recorders consist of three subsystems: an epitaxial GaAs-based optoelectronic chip, a Si CMOS IC, and a carbon fiber probe. Though this work encompasses the efforts of many, this dissertation outlines contributions in a few critical areas. To overcome low-flux LED emission, we devise an optical setup with ≈0.1% photon detection efficiency. Monte Carlo techniques model NIR scattering in biological tissue. Another steep challenge is the heterogeneous integration of the three material systems in a compact (200x170x150 um^3) package. To relay data and power between the GaAs and CMOS chips, through-wafer vias are critical. Using a novel selective copper plating technique, we demonstrate through-wafer GaAs vias with <2 Ohm series resistance. Additionally, conductive blind vias are presented for carbon fiber probe insertion. A self-aligned parylene etch mask permits sub-kOhm connection to a buried metal contact while maintaining GOhm substrate isolation. Both via structures meet the requirements of being low-resistance, insulated from the substrate, and amendable to thinned wafer processing. Finally, we demonstrate extensive processing on thinned chips and advances toward full heterogeneous integration via flip-chip alignment and solder bump bonding.