We focus on integrated photonics and nanophotonics, where a large number of subwavelength structures and/or optical functional components are densely integrated on a compact semiconductor chip (a small piece with typically a few millimeters square). By combining these cutting-edge devices with advanced information technology, we aim to harness the strengths of both optics and electronics, fully exploiting the unique features of light, such as ultrabroad bandwidth, high parallelism, and linearity, while relying on digital electronic circuits for complex control and signal processing. We target a variety of applications, including next-generation optical communication, sensing, and computing.
Followings are some examples of recent topics.
By using subwavelength structures (called metasurfaces) formed on a surface of a chip, we can freely manipulate the wavefront and polarization state of light. For example, with a properly designed elliptical nanopost array made of silicon on a silica chip, we can split incident light into its polarization components (Stokes vectors) and focus them onto six photodetectors. Through the judicious digital signal processing (DSP) of retrieved photocurrent signals, we have successfully demonstrated high-speed data transmission for the next-generation optical communication systems.
By embedding active electro-optic (EO) material inside a metasurface, we can dynamically modulate the intensity and wavefront of incident light at high speed. For example, we have proposed and demonstrated a novel light-modulating device that localizes normal-incident light inside a submicrometer-thick EO polymer film using a subwavelength grating structure. Such devices enable simultaneous modulation of highly parallelized optical signals as well as high-speed synthesis of optical wavefronts, which are desired for the future optical communication, interconnects, imaging, and computing.
Using the linearity and parallelism of light, various linear operations requied in the machine learning, such as multiply-accumulate (MAC) operations and convolutions, can be performed at the "speed of light" with minimal energy cosumption by simply transmitting light through an optical interferometric circuit. Such circuit, called "optical unitary processor (OUP)," has been actively researched over the past decade. We have proposed a novel OUP based on the concept of multiplane light conversion (MPLC) and demonstrated its excellent scalability over the conventional architectures. In addition to deep learning, such scalable OUPs are desired for optical quantum computing and space-division-multiplexed (SDM) optical communication.
Phased-array antennas used in wireless communications are applied to the optical wavelength band and integrated on few-millimeters-square semiconductor chips to realize high-speed beam steering and imaging. By controlling hundreds of optical phase shifters arranged in an array, arbitrary optical wavefronts can be synthesized, enabling, for example, high-speed switching of the emission direction. We have successfully demonstrated large-scale optical phased-array devices with over 100 channels as well as record-high-resolution beam steering by using our original concept of non-redundant array (NRA). These devices will be usefull for various imaging applications, including LiDAR (3D imaging sensors using laser light), which is indispensable for the future self-driving cars and autonomous robots. Novel computational imaging algorithms, such as single-pixel imaging and compressed sensing using these innovative devices, are also important research topics.