Advancing Life & Work
Di_Liang

Leading Silicon Chip Innovation From Santa Barbara

Although Silicon Valley may be located in the San Francisco Bay Area, technological breakthroughs can come from all places like the picturesque coastal city of Santa Barbara in southern California — a place that our photonics research group calls home.

 

More recently, we developed several critical building blocks, such as high-performance comb and microring lasers, high-speed modulators, and photodetectors on the same platform, plus another next-gen integration platform to enable a fully integrated optical link for future HPC applications. These foundational IP innovations will be key enablers for low-cost, low-latency, energy-efficient, and traffic-free data communication solutions in HPC and data centers.

 

About Our Photonics Research Group in Santa Barbara
Our mission is to innovate high-bandwidth, high-energy efficiency, and low-cost photonic solutions and foundational IPs that support and expand our business. We differentiate ourselves from other research groups because we develop novel materials, components, and special architecture, and we use a specialized integration technique to put different “secret sauces” on the same chip.

 

The Hewlett Packard Labs Photonics Team (photo taken pre-pandemic)The Hewlett Packard Labs Photonics Team (photo taken pre-pandemic)

 

Why Optical Communication Is Necessary for HPC
One of the unique advantages of optical communication over traditional electrical communication is in the same fiber, you can transmit a tremendous amount of data with different colors of light — that is, different wavelengths, so you can encode each wavelength with different data stream. Within the same fiber, you can securely transmit hundreds of channels of information without each channel interfering with each other over thousands of miles. That is impossible in electrical wire.

 

You need a high-speed connection between servers so they can talk to each other. Using traditional solutions like copper wire has already hit the physical limits of that technology. It doesn't provide enough bandwidth to transmit data and it consumes too much power, so an optical solution is a necessity.

 

The Memristor Laser
We just presented a paper with the title of “The Memristor Laser” in the prestigious IEEE International Electron Devices Meeting (IEDM). It is our most recent breakthrough to develop the world’s first memristor laser. It perfectly marries the merits of the memristor and diode laser to allow logic signals to be generated and stored in an optical fashion and in a non-volatile manner. It opens up a new landscape to extend our prowess from optical communications to optical and neuromorphic computing and beyond.

 

In the near term, we plan to show that this research can be used to combine computing, storage, and high-speed interconnect functionalities all together, which has never been done before. This integration of not only fundamental components but also functionalities in electrical and optical domains will greatly simplify the future HPC hardware architecture and chip design — leading to better energy efficiency, lower latency, smaller footprints, and an overall lower system cost.  We are all excited to venture this new research journey path in a better 2021.     

 

Looking Forward to 2021
In addition to continuing to develop our memristor-related photonic technology, we’re looking forward to integrating all key building blocks into a single transceiver chip, which we proposed in our ULTRALIT project funded by Advanced Research Projects Agency-Energy (ARPA-E) of the Department of Energy (DOE). It will be the first attempt in the world to build a fully integrated dense wavelength division multiplexing (DWDM) transceiver on silicon. Such a tiny transceiver chip at about 1/10 of a U.S. quarter coin in size can enable massive data communication with a capacity of over 1 terabits per second over a mile-long optical fiber and an energy efficiency of 1 pico Joule per bit or better. Such data transmission rate can support more than 125,000 people to stream HD (1080p) video at the same time.

 

Compared with current mainstream 100G transceiver products in the market, it will be a reduction of 100x in chip size and an improvement of 10x in data rate and 20x in power. This simultaneous, multi-dimensional improvement is a result of combining the optimal materials and device structures, a versatile photonic integration platform, and a novel DWDM transceiver architecture. Its success will not only pave the way to accomplish our multimillion-dollar program’s final milestone, but also will further demonstrate our tremendous technical leadership in next-gen integrated optical transceivers.

 

Of course, the true success enabler is long-term team innovation, dedication, and perseverance embodied by a team of passionate researchers in the lab, which I am always so grateful for and proud of.

 


Di Liang
Hewlett Packard Enterprise

twitter.com/hpe_labs
linkedin.com/showcase/hewlett-packard-labs/
labs.hpe.com

About the Author

Di_Liang

Di Liang is a senior research scientist for Hewlett Packard Labs in the Santa Barbara, leading a research group within the Large-Scale Integrated Photonics Lab. He was named a Fellow Member of The Optical Society in 2019. Liang holds a Bachelor of Science degree from Zhejiang University in China in Optical Engineering and a master’s and doctorate degree from University of Notre Dame in Electrical Engineering