Masaya Arahata | Searcharxiv

arXiv2026

Silicon photonic optical-electrical-optical converters based on load-resistor and current-injection operation

Optical-electrical-optical (OEO) converters are key primitives for low-latency, energy-efficient photonic computing because they enable nonlinear activation and optical signal regeneration on chip. We report two monolithically integrated silicon-photonic OEO converters-load-resistor (high-speed variant) and current-injection (high-gain variant) types-fabricated at a silicon photonics foundry. Each device combines a germanium photodetector with a micro-ring modulator (MRM). The converters exhibit reconfigurable nonlinear transfer functions and measurable on-chip RF OEO gain. The RF OEO gain scales linearly with the MRM bias power, with slopes of 0.10 mW^-1 (load-resistor of 10 kΩ) and 1.4 mW^-1 (current-injection), enabling a gain > 1 region at practical bias powers (~10 mW and ~1 mW, respectively). Eye diagrams confirm clear openings up to 4 Gb/s for a high-speed load-resistor variant with a 500-Ω load. To the best of our knowledge, this is the first experimental demonstration of a monolithically integrated, foundry-fabricated silicon-photonic load-resistor type OEO converter exhibiting reconfigurable nonlinear transfer and on-chip RF OEO gain. In the carrier-injection device, the activation slope exceeds unity, yielding 3.9 dB extinction-ratio regeneration. Short-pulse measurements yield 3-dB bandwidths of 1.49 GHz, 160 MHz (load-resistor of 500 Ω and 10 kΩ), and 76 MHz (current-injection), consistent with the RF data. Energy analysis shows an energy-bandwidth trade-off (RC-limited for load-resistor vs. lifetime-limited for injection) and outline routes to sub-pJ/bit operation via reduced capacitance and improved EO efficiency. These results establish silicon-photonic OEO converters as compact, foundry-compatible building blocks for scalable optoelectronic computing and optical neural networks.

physics.optics↗

arXiv2024

Optoelectronic recurrent neural network using optical-electrical-optical converters with RC delay

Optical neural network (ONN) has been attracting intense attention owing to their low latency and low-power consumption. Among the ONNs, optical recurrent neural network (RNN) enables low-power and high-speed time-series data processing using a compact loop structure. The loop losses need to be efficiently compensated so that the time-series information is maintained in the RNN operation. For this purpose, we focus on the optoelectronic RNN (OE-RNN) with optical-electrical-optical (OEO) converters to compensate for the loop losses. However, the effect of resistive-capacitive (RC) delay of OEO converters on the RNN performance is unclear. Here, we study in simulation an OE-RNN equipped with OEO converters with RC delay. We confirm that our modeled OE-RNN achieves the high training accuracy of time-series data classification even when RC delay is comparably large to the time interval of time-series data. Our analyses reveal that the accumulation of time-series data by RC delay does not degrade the RNN performance but rather can compensate for the degraded RNN performance due to loop losses. From the theoretical analysis referring to the gradient explosion and vanishing problems, we find the region related to loss and RC delay where the high training accuracy can be achieved. In simulation, we confirm this compensation effect in the large OE-RNN circuit up to 32$\times$32 scale. Our proposed scheme opens a new way of time-series data processing by utilizing RC delay for the optical computing and optical communication.

physics.optics↗