We are working on research on deep learning applications using programmable photonic circuits that integrate a large number of optical phase shifters on Si optical circuits. While the physical limits of transistor scaling are approaching and it is not possible to expect a dramatic improvement in the computing performance of CMOS electronic integrated circuits, the amount of computing required for artificial intelligence (AI) is increasing exponentially. Therefore, Innovative computing technology is required to realize highly advanced AI. To increase the computational capability without transistor scaling, photonic computing based on Si photonic integrated circuits is attracting more attentions. By using a programmable photonic integrated circuit in which a large number of optical phase shifters that electrically control the optical phase are integrated, arbitrary matrix operations can be performed in optical domain for deep learning and quantum computer. We have demonstrated an innovative hybrid optical phase shifter in which a III-V semiconductor membrane is bonded on a silicon photonic integrated circuit. By applying this hybrid optical phase shifter to programmable photonic integrated circuits, we are aiming to realize large-scale, ultralow power, and high-speed deep learning accelerators. We are challenging innovative computing by Si photonics from the viewpoint of devices, circuits, and systems.

In the AI/IoT era, enormous amounts of data communication will be required, while Moore's Law is coming to an end, and the I/O bandwidth of LSI is becoming saturated. In addition, signal delays and power consumption of electrical wiring in LSIs are becoming more serious, which hinders the improvement of LSI performance. In order to fundamentally solve these problems, we are working on research on optical communication devices based on Si photonics that realize LSIs with optical interfaces and optical wiring. In particular, we are conducting research on the hetrogeneous integration of Ge, III-V semiconductors, and 2D materials on Si platform. So far, we have investigated strained SiGe optical modulators, low dark current Ge photodetector, ultra-high performance hybrid optical modulators and Si/III-V hybrid. We have also proposed a III-V CMOS photonics platform that uses III-V semiconductor membrane on Si, enabling ultrahigh-performance electronic-photonic integrated circuits beyond Si photonics.

We are working on research on optical sensing using photonic integrated circuits. Various applications of optical sensing such as molecular sensing and radar are being studied, while most conventional optical sensing elements use spatial optical systems such as mirrors and gas cells, and there are major problems in terms of device size and cost. With the progress of Si photonics, research on photonic integrated circuits that integrate optical elements required for sensing on a single chip is underway all over the world. It is expected that photonic integrated circuits will realize compact, high-performance, and low-cost optical sensing. Our laboratory is investing new photonic integrated circuits for sensing that use germanium (Ge) instead of silicon. We aim to realize new optical sensing such as molecular sensors and medical / biosensors operating mid-infrared wavelengths.