Design and Optimization of a Hybrid Silicon–Silicon–Nitride Photonic Optical Digital-to-Analog Convertor (DAC) Enabled by Bi-State Phase-Change Materials

Abstract

As integrated photonic computing and optical signal-processing systems continue to advance, converting digital data into multi-level optical signals on chip is becoming a critical interface challenge. Conventional approaches often rely on electronic digital-to-analog converters (DACs) followed by arrays of optical modulators, increasing energy consumption, footprint, and overall latency. A photonic DAC addresses this bottleneck by directly mapping digital inputs to programmable optical transmission states, enabling compact and energy-efficient electronic-to-optical data generation. This work presents the design and simulation of a novel hybrid silicon–silicon-nitride (Si–SiN) photonic DAC, called HyPoD, based on bi-state phase-change-material (PCM) germanium antimony telluride (Ge2Sb2Te5, GST) cells integrated on silicon-on-insulator waveguides. HyPoD encodes digital inputs as discrete analog optical power levels by exploiting the nonvolatile, electrically driven switching of multiple GST cells between fully amorphous and fully crystalline states using doped-silicon microheaters. By tailoring the GST cell lengths, the fully crystalline state introduces binary-weighted optical losses of 1, 2, 4, and 8 dB, while the amorphous state provides nearly transparent transmission with ≈0 dB loss. Selectively switching the four GST cells therefore enables 16 programmable output optical power levels, corresponding to four-bit operation, across a 15-dB optical dynamic range. As a case study, we further evaluate the energy impact of integrating HyPoD with a photonic matrix-vector multiplier and identify the resulting system-level performance trade-offs.