Optimizing Thermal Dynamics and Multilevel Control in Silicon Doped Microheaters for Optical Phase Change Memory

Nonvolatile memory (NVM) offers advantages for high performance computing such as improved energy efficiency and fast switching speeds. One approach is to use Phase Change Materials (PCM) with photonic integrated circuits. Both optical and electric approaches have been studied to switch the PCM from a high resistance state (amorphous) to a low resistance state (crystalline). Optical switching enables accurate multilevel control but is limited in scalability. In contrast, electrical switching offers effortless CMOS integration but faces reliability challenges to achieve reversible control. Recent studies using microheaters, for indirect switching, have demonstrated an inherent tradeoff between scalability and multilevel control. To address this hypothesis, two heater designs have been evaluated, metal and silicon doped microheaters. While metal heaters can efficiently offer a larger heating area, they require a thicker oxide spacing between the heater and substrate to avoid optical interference. Understanding the heat dissipation in silicon doped is thus vital for the optimization to accurately control the PCM. Addressing challenges such as high PCM melting temperatures, and rapid quenching rates required for re-amorphization is critical for achieving precise multilevel control. Various heater designs are investigated to address these issues. First, a bowtie design reveals challenges with multi-level accuracy as it relies on the stochastic nucleation process for intermediate state modulation. In contrast, a five bridge (equally spaced with same width) design controls the PCM’s state through precise spatial temperature control at each island hotspot. However, since each hotspot simultaneously reaches a similar temperature, the control of intermediate multilevel states is limited. This study experimentally assesses an alternative multi-bridge design that varies each bridge width (non-homogeneous hotspots) and results in an overall triangle-like temperature distribution to achieve continuous multilevel control. <br/>Previous efforts have characterized the steady state temperature distribution of microheaters using thermoreflectance techniques and numerical models. Under normal operating conditions, however, the microheaters are biased with sub microsecond pulse widths to achieve higher power densities that can reach the crystallization temperature of the PCM. Advanced thermal characterization techniques with both high spatial (&lt; 1 µm) and temporal resolution (&lt; 0.5 µs) are thus needed to quantify the impact of the microheater design on its transient thermal dynamic behavior. Specifically, the dependence of the electrical pulse width on the peak temperature, energy consumption, and the quenching rate has not been experimentally verified. This work leverages high throughput CCD based Transient Thermoreflectance Imaging with high speed pulsed IV to assess the design parameters on the thermal response of multi-bridge microheaters. <br/>First, the microheater’s transient thermal rise/decay was analyzed to extract the thermal time constants and obtain quenching rates of ≈ 0.8 K/ns. Under steady state conditions, temperature scales linearly with power. Based on the magnitude of the thermal time constants (≈ 0.48 µs), however, the maximum temperature was achieved in the non-linear regime where the tradeoff between maximum power density and transient heat accumulation must be assessed. This study reveals the maximum power density and temperature was achieved by reducing the pulse width down to 0.2 µs. As the energy consumption is inversely proportional to the electrical pulse width, the minimum energy was also achieved at 0.2 µs. Overall, short pulses enable effective management of higher temperatures, smaller quenching rates, and lower energy consumption per cycle of crystallization and amorphization processes.