The next generation of wireless systems may be built on microjet-cooled glass
Wireless communications like cellphone and Wi-Fi networks are everywhere in the 21st century, continually increasing in size, speed, and complexity. But energy efficiency isn’t keeping up. By some estimates, the wireless ecosystem as a whole may be responsible for 23% of all global CO2 emissions by 2030.
Purdue University researchers may have a solution. They are collaborating on a $1.5 million grant from the National Science Foundation (NSF) on heterogenous integration of the high-power electronics involved, built on a glass substrate. They are also developing novel cooling methods to decrease the energy consumed by these next-generation wireless networks.
“The wireless spectrum is getting very crowded very quickly,” said Tiwei Wei, assistant professor of mechanical engineering. “So the newest 5G communication standards use something called millimeter wave technology, which offers more bandwidth. But it also requires a lot more power on the back end.”
These high-performance systems require a different kind of semiconductor, referred to as a wide bandgap semiconductor, whose power density is ten times that of current silicon-based systems. As engineers attempt to consolidate these devices on their backend equipment — antennas, semiconductors, power amplifiers, and transmitters — a tremendous amount of heat is generated, which leads to impacted performance and equipment failures.
That’s where Purdue comes in. An acknowledged leader in semiconductor research, Purdue already participates in an NSF program investigating the future of heterogenous integration of computer chips. Wei’s lab focuses specifically on semiconductor packaging and cooling, especially for high-performance applications like servers and data centers. This made him the perfect choice to tackle the thermal challenges of these next-generation wireless systems.
“There are many cooling technologies for high-power chips,” said Wei. “For this application, we are investigating embedded microjet cooling using glass packaging fabrication technologies, which involves tiny fluid channels built into the chips themselves. This is critical because in heterogenous integration, the chips are stacked on top of each other, so the cooling systems must be built in.”
Wei is part of a team who are working together on a multi-prong approach: designing the circuity, heterogeneously integrating the chips, cooling the chips, and using machine learning to optimize the power amplification of the transmitters. Kenle Chen, assistant professor of electrical engineering at University of Central Florida, is the principal investigator, and will work on the circuit design for the high-frequency system. Co-principal investigator Zheng Zhang, associate professor of electrical and computer engineering at University of California Santa Barbara, is responsible for the machine learning aspect of the project.
Another novel aspect is the use of a glass substrate in building the physical chip architecture. “Glass is the perfect material in high-frequency applications like this, because it has excellent electrical properties, good thermal and mechanical stability, and low signal crosstalk,” said Wei. “But there are many thermal challenges to using glass in high-power and high-performance systems. That’s why this team reached out to me, because I have worked with cooling glass chip stacks since 2014.”
For this 4-year $1.5 million project, Wei will be building physical prototypes at Purdue’s Birck Nanotechnology Center. This will be the first time that glass packaging wafers will be manufactured in Birck, which traditionally has focused on silicon.
“I’m so excited for this project,” said Wei. “We are looking forward to seeing what our cooling technology can accomplish when put together with their next-generation communications systems.”
Writer: Jared Pike, jaredpike@purdue.edu, 765-496-0374
Source: Tiwei Wei, tiwei@purdue.edu