Reflective Multimode Interference Coupler

Multimode interference (MMI) couplers are fundamental components in Photonic Integrated Circuits (PICs), enabling compact interconnects between modulators, filters, detectors, and other optical devices on a single chip. Conventional MMI couplers, however, tend to be relatively long, creating challenges for efficient integration in dense PIC layouts. To overcome these limitations, researchers have proposed a compact reflective MMI coupler featuring etched Total Internal Reflection (TIR) mirrors, a design that reduces device footprint while improving integration potential.

Accurate simulation of this reflective MMI structure is non‑trivial due to the limitations of individual modeling methods. FullWAVE FDTD struggles with the overall size of the MMI region, especially when three‑dimensional accuracy is required, while BeamPROP BPM cannot account for strong facet reflections produced by the TIR mirrors. RSoft Photonic Device tools address these challenges with a suite of solvers tailored to specific photonic problems. By decomposing the device into the MMI body, the reflective facets, and the optimization stages, engineers can apply the most efficient simulation method to each part.

This application note presents a co‑simulation workflow combining BeamPROP, FullWAVE, and MOST to efficiently model the reflective MMI coupler. The process begins with BeamPROP BPM to simulate forward light propagation through the MMI region, leveraging BPM’s speed and suitability for large structures. The output field from this step is exported to FullWAVE FDTD, which accurately models the omnidirectional reflection behavior at the etched TIR mirror facets. FullWAVE captures the reflected field, which is subsequently fed back into BeamPROP to model backward propagation and assess reflection efficiency at the input ports.

The final stage utilizes MOST, RSoft’s optimization tool, to fine‑tune the MMI length for maximum efficiency. MOST performs multi‑variable optimization using user‑defined scripts and simulation outputs. This integrated simulation flow makes it possible to identify an optimal MMI length of approximately 34.7 µm while achieving a high reflection efficiency of around 94.9%.

A major advantage of this co‑simulation method is its dramatic reduction in computing time and memory requirements compared to running a full‑device FDTD simulation alone. A FullWAVE‑only simulation of the entire reflective MMI structure requires approximately 20 hours on an 8‑core system with 20 GB of RAM, whereas the combined BeamPROP + FullWAVE approach completes within about 10 minutes using only 1.5 GB of RAM. This efficiency gain demonstrates the value of applying the right solver for each part of the device physics problem.

Overall, this co‑simulation strategy successfully models the reflective MMI coupler by leveraging BPM for long‑range propagation, FDTD for localized facet reflections, and MOST for device optimization. The method provides a highly efficient workflow for simulating compact photonic components that are otherwise too complex or resource‑intensive for a single solver. This approach enables faster design cycles and supports the development of more tightly integrated PICs, addressing the growing demand for compact, high‑performance photonic systems.