A magneto-optical isolator is demonstrated for use with a Si waveguide. The isolator is based on a Mach–Zehnder interferometer employing a nonreciprocal phase shift and is fabricated by bonding a magneto-optic garnet CeY2Fe5O12 (Ce:YIG) directly onto the Si waveguide. The surface-activated bonding is based on oxygen-plasma exposure in a high-vacuum chamber. The nonreciprocal phase shift is observed by applying an external magnetic field. An isolation ratio of 21dB is obtained at a wavelength of 1559nm.

Silicon waveguide optical non-reciprocal devices based on the magneto-optical effect are reviewed. The non-reciprocal phase shift caused by the first-order magneto-optical effect is effective in realizing optical non-reciprocal devices in silicon waveguide platforms. In a silicon-on-insulator waveguide, the low refractive index of the buried oxide layer enhances the magneto-optical phase shift, which reduces the device footprints. A surface activated direct bonding technique was developed to integrate a magneto-optical garnet crystal on the silicon waveguides. A silicon waveguide optical isolator based on the magneto-optical phase shift was demonstrated with an optical isolation of 30 dB and insertion loss of 13 dB at a wavelength of 1548 nm. Furthermore, a four port optical circulator was demonstrated with maximum isolations of 15.3 and 9.3 dB in cross and bar ports, respectively, at a wavelength of 1531 nm.

We demonstrate a low-crosstalk 2 × 2 thermo-optic switch with silicon wire waveguides. The device is based on a 2 × 2 array of Mach–Zehnder interferometer (MZI) switches. Lowest crosstalk levels of –50 dB and –30 dB are obtained for ‘bar’ and ‘cross’ switching states, respectively. An intersection in the switch is important for low-crosstalk operation. The power consumption of one MZI element switch is 40 mW and the total power consumption of the device is at most 160 mW.

We report a multi-mode interference-based optical gate switch using a Ge₂Sb₂Te₅ thin film with a diameter of only 1 µm. The switching operation was demonstrated by laser pulse irradiation. This switch had a very wide operating wavelength range of 100 nm at around 1575 nm, with an average extinction ratio of 12.6 dB. Repetitive switching over 2,000 irradiation cycles was also successfully demonstrated. In addition, self-holding characteristics were confirmed by observing the dynamic responses, and the rise and fall times were 130 ns and 400 ns, respectively.

We, for the first time, present the ultrafast optical nonlinear response of a hydrogenated amorphous silicon (a-Si:H) wire waveguide using femtosecond pulses. We show cross-phase and cross-absorption modulations measured using the heterodyne pump-probe method and estimate the optical Kerr coefficient and two-photon absorption coefficient for the amorphous silicon waveguide. The pumping energy of 0.8 eV is slightly lower than that required to achieve two-photon excitation at the band gap of a-Si:H (~1.7 eV). An ultrafast response of less than 100 fs is observed, which indicates that the free-carrier effect is suppressed by the localized states in the band gap.

Processing information in the optical domain promises advantages in both speed and energy efficiency over existing digital hardware for a variety of emerging applications in artificial intelligence and machine learning. A typical approach to photonic processing is to multiply a rapidly changing optical input vector with a matrix of fixed optical weights. However, encoding these weights on-chip using an array of photonic memory cells is currently limited by a wide range of material- and device-level issues, such as the programming speed, extinction ratio and endurance, among others. Here we propose a new approach to encoding optical weights for in-memory photonic computing using magneto-optic memory cells comprising heterogeneously integrated cerium-substituted yttrium iron garnet (Ce:YIG) on silicon micro-ring resonators. We show that leveraging the non-reciprocal phase shift in such magneto-optic materials offers several key advantages over existing architectures, providing a fast (1 ns), efficient (143 fJ per bit) and robust (2.4 billion programming cycles) platform for on-chip optical processing.