Search for: All records

Scalable programmable photonic integrated circuits (PICs) can potentially transform the current state of classical and quantum optical information processing. However, traditional means of programming, including thermo-optic, free carrier dispersion, and Pockels effect result in either large device footprints or high static energy consumptions, significantly limiting their scalability. While chalcogenide-based non-volatile phase-change materials (PCMs) could mitigate these problems thanks to their strong index modulation and zero static power consumption, they often suffer from large absorptive loss, low cyclability, and lack of multilevel operation. Here, we report a wide-bandgap PCM antimony sulfide (Sb₂S₃)-clad silicon photonic platform simultaneously achieving low loss (<1.0 dB), high extinction ratio (>10 dB), high cyclability (>1600 switching events), and 5-bit operation. These Sb₂S₃-based devices are programmed via on-chip silicon PIN diode heaters within sub-ms timescale, with a programming energy density of$$\sim 10\,{fJ}/n{m}^{3}$$$~10fJ/n{m}^3$. Remarkably, Sb₂S₃is programmed into fine intermediate states by applying multiple identical pulses, providing controllable multilevel operations. Through dynamic pulse control, we achieve 5-bit (32 levels) operations, rendering 0.50 ± 0.16 dB per step. Using this multilevel behavior, we further trim random phase error in a balanced Mach-Zehnder interferometer.

Programmable photonics play a crucial role in many emerging applications, from optical accelerators for machine learning to quantum information technologies. Conventionally, photonic systems are tuned by mechanisms such as the thermo-optic effect, free carrier dispersion, the electro-optic effect, or micro-mechanical movement. Although these physical effects allow either fast (>100 GHz) or large contrast (>60 dB) switching, their high static power consumption is not optimal for programmability, which requires only infrequent switching and has a long static time. Non-volatile materials, such as phase-change materials, ferroelectrics, vanadium dioxide, and memristive metal oxide materials, can offer an ideal solution thanks to their reversible switching and non-volatile behavior, enabling a truly “set-and-forget” programmable unit with no static power consumption. In recent years, we have indeed witnessed the fast adoption of non-volatile materials in programmable photonic systems, including photonic integrated circuits and free-space meta-optics. Here, we review the recent progress in the field of programmable photonics, based on non-volatile materials. We first discuss the material’s properties, operating mechanisms, and then their potential applications in programmable photonics. Finally, we provide an outlook for future research directions. The review serves as a reference for choosing the ideal material system to realize non-volatile operation for various photonic applications.

Embedding is widely used in recommendation models to learn feature representations. However, the traditional embedding technique that assigns a fixed size to all categorical features may be suboptimal due to the following reasons. In recommendation domain, the majority of categorical features' embeddings can be trained with less capacity without impacting model performance, thereby storing embeddings with equal length may incur unnecessary memory usage. Existing work that tries to allocate customized sizes for each feature usually either simply scales the embedding size with feature's popularity or formulates this size allocation problem as an architecture selection problem. Unfortunately, most of these methods either have large performance drop or incur significant extra time cost for searching proper embedding sizes. In this article, instead of formulating the size allocation problem as an architecture selection problem, we approach the problem from a pruning perspective and proposePruning-basedMulti-sizeEmbedding (PME) framework. During the search phase, we prune the dimensions that have the least impact on model performance in the embedding to reduce its capacity. Then, we show that the customized size of each token can be obtained by transferring the capacity of its pruned embedding with significant less search cost. Experimental results validate that PME can efficiently find proper sizes and hence achieve strong performance while significantly reducing the number of parameters in the embedding layer.

Increasing the space-bandwidth product of spatial light modulators incurs severe issues in terms of power consumption, mutual crosstalk, and control signal wiring. In this opinion article, we propose a novel system to overcome these challenges by marrying energy-efficient modulators in photonic integrated circuits (PICs) and a meta-optical beam aggregator. This hybrid approach can significantly improve the space-bandwidth product, theoretically up to 10¹³Hz · pixel, which is several orders of magnitude higher than the state-of-the-art.