SBIR Phase I: Tunable 1550-nm VCSEL Integrated on Silicon-Photonics Platform for Energy-efficient Broadband Data Communications

该小型企业创新研究(SBIR)第一阶段项目将解决数据中心内距离长达2公里的高带宽(100+Gb/S)、低功率、低成本的波分复用(WDM)光通信链路问题。这些链接是下一代互联网/云/超级计算应用程序所必需的。具体地说，这个项目将检查激光器，它们构成了成本和功率需求的大部分。该项目将通过改进现有的低成本、低功率的激光器结构-垂直腔面发射激光器(VCSEL)来解决这一问题，以适应非制冷WDM操作。具体地说，本项目将研究可调谐垂直腔面发射激光器及其与高对比度光栅(HCG)光耦合器的组合，该耦合器可以将表面正常的垂直腔面发射激光器以非常高的效率耦合到面内波导。该芯片将包括一个直接调制的波长可调的VCSEL和一个表面法线到面内的波导耦合器，它可以扩展成WDM阵列，用于将VCSEL阵列多路复用到一根单模光纤中。第一阶段的目标包括可调谐垂直腔面发射激光器的理论模型和设计工具，以及适用于WDM应用的高效耦合器设计，以及10Gbps直接调制1550 nm垂直腔面发射激光器的实验演示和表征。该项目更广泛的影响/商业潜力是大幅降低数据中心和超级计算机内部光纤链路的成本和能源需求。去年，谷歌和其他以数据为中心的大型公司一直在要求购买这类产品，以保持其数据中心设施的增长速度。目前使用DFB激光器的WDM激光阵列解决方案需要10倍于基于VCSEL的方法的功率和10倍的成本。另一方面，目前基于VCSEL的850 nm链路在没有耗电的TEC冷却器的情况下无法制成WDM源，因为它们缺乏精确的网格化波长控制。此外，使用目前的器件结构，它们不能容易地组合成单模光纤，因此它们不能被多路复用。使用可调谐的VCSEL解决了栅格波长控制的问题，并且高效的耦合器可以在没有显著损耗的情况下组合激光输出。在WDM系统中实现可调VCSEL将使数据中心的高速光纤链路的成本和能源需求降低到原来的1/10，从而使下一代数据中心和超级计算机应用的计算能力进一步扩展。

This Small Business Innovation Research (SBIR) Phase I project will address the problem of high bandwidth (100+ Gb/s), low power, low cost wavelength division multiplexed (WDM) optical communications links of distances up to 2 km inside of data centers. These links are necessary for next generation internet/cloud/supercomputing applications. Specifically, this project will examine the lasers, which make up much of the cost and power requirements. The project will address the problem by improving an existing low cost, low power laser structure, vertical cavity surface emitting lasers (VCSELs), to be suitable for uncooled WDM operation. Specifically, this project will study tunable VCSELs and their combination with a high contrast grating (HCG) optical coupler, which can couple surface-normal VCSELs to an in-plane waveguide with very high efficiency. The chip will include a directly modulated wavelength-tunable VCSEL and a surface-normal to in-plane waveguide coupler, which can be extended into an array for WDM multiplexing an array of VCSELs into one single-mode fiber. The Phase I goals include a theoretical model and design tools for tunable VCSELs with efficient coupler design suitable for WDM applications, and experimental demonstration and characterization of a 1550-nm VCSEL with 10 Gbps direct modulation. The broader impact/commercial potential of this project is a drastic reduction in the cost and energy requirements of optical links inside of data centers and supercomputers. Companies such as Google and other large data-centric companies have been asking for this type of product in the last year to maintain the rate of growth in their data center facilities. Present WDM laser array solutions using DFB lasers require 10X the power and 10X the cost of a VCSEL-based approach. Present 850-nm VCSEL-based links, on the other hand, cannot be made into a WDM source without power-hungry TEC coolers due to their lack of precise, gridded, wavelength control. Additionally they cannot easily be combined into a single mode fiber using present device structures, so they cannot be multiplexed. Using a tunable VCSEL solves both the problem of gridded wavelength control, and an efficient coupler can combine the laser outputs without significant loss. The realization of tunable VCSELs in WDM systems will result in a 10X reduction in both cost and energy requirements in high-speed optical links for data centers, enabling the further scaling of computational power for next generation data center and supercomputer applications.