International Scientific Journals
of Scientific Technical Union of Mechanical Engineering "Industry 4.0"

Traditional 1 × 2 multimode interference (MMI) splitters often encounter challenges such as high back reflections and limited output flexibility, which typically require additional structures like tapers or S-bends. These constraints limit their performance in advanced photonic networks. To overcome these issues, we propose an AI-optimized angled multimode interference (AMMI) splitter featuring a buried silicon nitride (SiN) core with silica cladding. By employing an angled propagation path, the device minimizes reflections to the source and allows greater adaptability for waveguide interconnections in dense photonic circuits. The design optimization was performed using a combination of artificial intelligence (AI) algorithms integrated with full-vectorial beam propagation method (FV-BPM) and finite-difference time-domain (FDTD) simulations. AI-driven parameter scanning enabled efficient exploration of the design space, improving device performance and robustness compared to manual optimization. The proposed AMMI splitter achieves an excess loss of 0.22 dB and an output imbalance of 0.001 dB at 1.31 μm, with total device length of 101 μm and thickness of 0.4 μm. Over the full O-band (1260–1360 nm), performance remains stable, with excess loss below 1.57 dB and imbalance below 0.05 dB, while maintaining back reflections as low as –40 dB. The compact CMOS-compatible design demonstrates high tolerance to fabrication deviations, making it highly suitable for large-scale integration. With its AI-enhanced optimization process, the proposed splitter supports high-speed, low-loss transmission for O-band photonic networks and data-center interconnects, offering scalability and reliability for next-generation optical systems.

Back reflection challenges significantly constrain the efficiency of optical communication networks utilizing dense wavelength division multiplexing (DWDM) technology based on silicon multimode interference (MMI) waveguides. To address this issue, we propose an innovative 1×4 optical demultiplexer design based on MMI within a silicon-nitride (SiN) strip waveguide configuration that operates within the C-band spectrum. Our simulation outcomes indicate that the proposed device efficiently transmits four channels with 10 nm spacing in the C-band, exhibiting low power loss ranging from 1.99-2.36 dB, extensive bandwidth of 7.69-8.09 nm, and good crosstalk values between 20.7-23.5 dB. Utilizing the low refractive index of SiN, we achieve exceptionally low back reflection of 40.58 dB without requiring specialized angled MMI designs, typically needed in Si MMI technology. Hence, this SiN-based MMI demultiplexer technology can be effectively employed in DWDM systems to achieve high data transfer rates with minimal back reflection in optical communication systems

In this paper, we introduce an innovative design for an all-optical half adder (HA) utilizing a pair of dual-ring resonators within a 2-dimensional square lattice photonic crystal (PC) framework, all achieved without the use of nonlinear materials. The all-optical HA encompasses both AND and XOR gates, each constructed with cross-shaped waveguides and twin ring resonators positioned in a 2D square lattice PC. These resonators are filled with silicon (Si) rods set in a silica (SiO2) medium. Employing the plane-wave expansion (PWE) and finite difference time domain (FDTD) techniques, we comprehensively analyzed and simulated the behavior of the AND and XOR gates. The simulation outcomes reveal that the internal light propagation within the device emulates the functions of traditional AND and XOR gates.
Consequently, the proposed device displays promising potential for integration into optical arithmetic logic units, thereby enhancing digital
computing circuits.
The structural configuration encompasses an optical AND gate and an optical XOR gate, meticulously designed to operate within the Cband spectrum. The results unequivocally demonstrate a distinct demarcation between logic states 1 and 0, encapsulating a narrow power range that contributes to heightened robustness and minimized logic errors within the photonic decision circuit. Hence, the proposed HA stands as a pivotal building block in the creation of cutting-edge photonic arithmetic logic units.