Scholarlink

<p>Artificial intelligence, machine learning and high-performance computing workloads are pushing electrical input/output to its limits in signal reach, energy efficiency and bandwidth density, turning optics from option to necessity. Complementary metal–oxide–semiconductor-integrated silicon photonics offers a practical path forward by combining high-volume manufacturing with mature photonic building blocks. This Review presents progress across devices (on-chip lasers and semiconductor optical amplifiers, compact modulators, high-speed photodetectors, low-loss routing and efficient chip–fibre couplers), multimaterial integration (hybrid assembly, heterogeneous wafer bonding, microtransfer printing and monolithic epitaxy) and electronics co-design (digital signal processing, serializer/deserializer, stacked-driver topologies, bias control and thermal tuning) to show how total link energy is being driven towards the sub-picojoule per bit regime. We connect these advances to system architectures that are evolving from pluggables to linear-drive pluggables and co-packaged optics, and we discuss the trade-offs among bandwidth density, thermal design, yield and cost. We identify near-term bottlenecks, notably thermal pathways and manufacturing yield, and highlight technologies most likely to unlock the next jump in performance, including on-chip comb sources for dense wavelength-division multiplexing and wafer-scale 3D electronic and photonic stacks. The same platform is poised to impact optical compute input/output, sensing and quantum photonics, linking device-level innovation to system-level gains across computing and communications.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrating silicon photonics with complementary metal–oxide–semiconductor technologies

  • 期刊论文

  • 发表日期：

    2025年11月7日

查看全文
  • Yating Wan,
  • William He,
  • James Jaussi,
  • Ling Liao,
  • David Z. Pan,
  • John E. Bowers,
  • Haisheng Rong

摘要

Artificial intelligence, machine learning and high-performance computing workloads are pushing electrical input/output to its limits in signal reach, energy efficiency and bandwidth density, turning optics from option to necessity. Complementary metal–oxide–semiconductor-integrated silicon photonics offers a practical path forward by combining high-volume manufacturing with mature photonic building blocks. This Review presents progress across devices (on-chip lasers and semiconductor optical amplifiers, compact modulators, high-speed photodetectors, low-loss routing and efficient chip–fibre couplers), multimaterial integration (hybrid assembly, heterogeneous wafer bonding, microtransfer printing and monolithic epitaxy) and electronics co-design (digital signal processing, serializer/deserializer, stacked-driver topologies, bias control and thermal tuning) to show how total link energy is being driven towards the sub-picojoule per bit regime. We connect these advances to system architectures that are evolving from pluggables to linear-drive pluggables and co-packaged optics, and we discuss the trade-offs among bandwidth density, thermal design, yield and cost. We identify near-term bottlenecks, notably thermal pathways and manufacturing yield, and highlight technologies most likely to unlock the next jump in performance, including on-chip comb sources for dense wavelength-division multiplexing and wafer-scale 3D electronic and photonic stacks. The same platform is poised to impact optical compute input/output, sensing and quantum photonics, linking device-level innovation to system-level gains across computing and communications.