GASP: Gallium Arsenide (III-V) photonic integrated circuits built like Silicon Photonics

在技术领域，很少有指数趋势能长期保持指数趋势。摩尔定律可能是最好的例子，它的指数增长（现代微处理器中晶体管的数量每18-24个月翻一番）持续了近40年，给现代社会带来了巨大的好处。另一方面，近年来，出现了另一个稍微令人担忧的指数。这是我们作为一个社会一直在消耗的数据总量。在过去的十年里，它一直呈指数级增长，没有任何放缓的迹象。现代数据中心的规模和数量在全球范围内都在增长，到2030年，它们预计将消耗世界总电力供应的20%，甚至在今天，它们排放的二氧化碳比全球航空业还多。如果你更仔细地观察一个数据中心，你会发现大部分能量不是在计算上消耗的，而是在以非常高的数据速率在可变距离上发送比特时消耗的，而这主要是在光学领域完成的。如果我们能建立一个更有效的光子集成电路来处理这种光通信，我们就能在原则上解决这个能量问题。作为该项目的一部分，所做的工作旨在开发基础平台和可扩展的制造过程，以构建这些高效的光子器件。我们的方法是将最好的制造工艺（来自硅光子学和硅MEMS代工厂）应用于最好的光学材料（化合物半导体）。我们相信这是构建最节能的集成光子器件的自然途径。除了数据中心收发器，这里开发的平台也将适用于其他领域，从用于卫星通信的光子设备到用于量子信息的低温光子平台。

There are very few exponential trends in technology that stay exponential for long. Moore's law is probably the best example, where an exponential (the doubling of the number of transistors in modern microprocessors every 18-24 months) has persisted for almost four decades, with great benefit to modern society. On the other hand, in recent years, there has been another slightly more worrying exponential. This is the total amount of data that we as a society have been consuming. It has been growing exponentially for the past decade and shows no signs of slowing down. It is probably best illustrated by modern data centers that have grown in size scale and number all around the globe to the point where by 2030, they are expected to consume ~ 20% of the world's total electricity supply and even today, they emit more CO2 than the global airline industry. If you look at a data center more carefully, most of the energy is dissipated not in computing, but in sending bits around at very high data rates over variable distances, and this is predominantly done in the optical domain.If we can build a more efficient photonic integrated circuit for handling this optical communication, we can address this energy problem in principle. The work done as part of this project is aimed towards developing the underlying platform and a scalable manufacturing process for building these efficient photonic devices. Our approach is to apply the best manufacturing processes (derived from silicon photonics and silicon MEMS foundries) to the best available optical materials (compound semiconductors). We believe this is a natural route towards building the most energy efficient integrated photonic devices. In addition to data center transceivers, the platform developed here will also be applicable to other areas ranging from photonic devices for satellite communication to cryogenic photonics platforms for quantum information.