Kyocera and Tohoku University Integrate Optical Isolator Directly on Silicon Photonics

A locally applied laser-annealing process enabled the researchers to crystallise magneto-optical garnet without exposing the entire photonic chip to damaging temperatures. The experimental device achieved an isolation ratio of 13.6 dB.

Kyocera and Tohoku University have demonstrated an optical isolator fabricated directly on a silicon-photonics circuit, using localised laser annealing to overcome a significant thermal-integration problem.

Optical isolators allow light to propagate in one direction while suppressing reflections travelling back towards the laser source. These reflections can destabilise a laser and impair the performance of optical communication systems, making isolation an important function in increasingly dense photonic circuits.

Today, isolators are commonly implemented as separate optical components. Integrating this function directly on a silicon-photonics chip could reduce component count, assembly complexity and optical coupling losses in future optical engines and co-packaged optics systems.

Local heating addresses the thermal-budget problem

The device developed by Kyocera and Tohoku University uses magneto-optical iron garnet deposited on a silicon waveguide. The material must be crystallised at approximately 600°C or higher to obtain the properties required for non-reciprocal optical transmission.

Applying that temperature to the entire chip can damage metal electrodes, interconnects and other structures already fabricated on the photonic circuit. This incompatibility with the chip’s thermal budget has been one of the obstacles to monolithic integration of garnet-based isolators.

Instead of heating the complete wafer or die in a furnace, the researchers applied a near-infrared laser to an area measuring approximately 700 × 700 µm. The process concentrates the thermal treatment in the isolator region while limiting heat exposure elsewhere on the chip.

Electron-microscope observations subsequently confirmed that the laser-treated garnet had crystallised on the silicon waveguide.

Experimental device delivers 13.6 dB isolation

The resulting device is a magneto-optical Mach-Zehnder isolator. Its operation relies on interference between optical paths and a non-reciprocal phase shift introduced by the magneto-optical material.

Testing in the optical-communications wavelength range produced an isolation ratio of 13.6 dB. According to the research team, this corresponds to suppressing approximately 95% of the light reflected towards the source.

The result demonstrates that local laser annealing can create a functioning optical-isolation structure directly on a silicon-photonics circuit. It also avoids the need to subject surrounding circuitry to the garnet’s full crystallisation temperature.

This integration route could become relevant to co-packaged optics, where optical interfaces are placed close to processors, switches or other high-bandwidth electronic devices. Shorter electrical paths can reduce I/O power, but denser optical integration also makes the behaviour of lasers, waveguides and reflected signals more critical.

Manufacturing questions remain open

The demonstration is an important process-development milestone, but it is not yet evidence of a production-ready optical isolator.

Neither Kyocera nor Tohoku University has published the device’s insertion loss, operating bandwidth, polarisation dependence, temperature stability or maximum supported optical power. These parameters will determine whether the isolator can be used in practical optical transceivers and co-packaged optical engines.

Manufacturing information is also limited. The researchers have not reported process yield, wafer-level uniformity, alignment tolerances, long-term reliability or the time required to anneal each isolator. Local processing must ultimately be sufficiently repeatable and productive to compete with separately manufactured optical components.

Kyocera and Tohoku University say their next objectives include reducing optical loss, improving efficiency and developing the productivity needed for mass manufacturing.

The work was published in IEEE Access under the title “Monolithic Magneto-Optical Mach-Zehnder Isolator Using Laser-Annealed Iron Garnet on a Silicon Waveguide.”

For silicon-photonics developers, the result is therefore best viewed as a demonstrated solution to a specific integration problem: crystallising a high-temperature magneto-optical material after sensitive structures are already present on the chip. Whether the approach can satisfy the optical-performance and manufacturing requirements of commercial co-packaged optics remains the next engineering test.