MIT and Ericsson Collaborates to Research New Generation of Energy-Efficient Computing Networks
Researchers from MIT and Ericsson have collaborated on two research projects focusing on solving energy demands of future computing networks.
Silicon-based computational systems today are reaching their limits of computational speed, performance, and other factors. Due to the ever-decreasing feature size of circuit elements such as transistors, the number of devices on a single chip has increased exponentially. As a result of this high device density, the energy required by the systems is significant. On the other hand, such highly integrated systems are necessary for complex computational tasks like applying artificial intelligence (AI) and machine learning (ML) models.
Power-intensive AI programs are becoming common in industries. Due to their ability to handle high volumes of data, AI systems are giving Industry 4.0 a significant boost. Intelligent AI-led solutions identify trends and patterns that can be used to make the manufacturing processes more efficient and reduce their energy consumption. Apart from industries, the influence of AI can be seen across sectors such as transportation, education, e-commerce, communication, robotics, and healthcare. In addition, AI is transforming future technologies such as autonomous vehicles and virtual healthcare systems.
With these trends, we are talking about a dramatic increase in device densities on chips. With silicon-based chips, the energy consumption will rise exponentially if these chips are scaled for such power-intensive applications.
To solve these problems of energy demands, the Massachusetts Institute of Technology (MIT) and Ericsson have collaborated on two distinct research projects. The first research focuses on the discovery of new potential materials to implement computer chips that are inspired by the human brain and use significantly less energy than existing silicon-based chip designs.
The second project aims to make electronic systems truly autonomous by removing the need for charging. The researchers look to power devices from ambient radio frequency waves conventionally used for TV and communication signals.
Lithium is a key element in lithium batteries, which are used in applications such as electric vehicles. However, MIT professors Jennifer Rupp and Martin Bazant have recently indicated that lithium-based battery electrodes are also ideal for other applications, including computing. One of the examples of such applications is that lithium oxide materials could be a fundamental component for memristors, alternatives of transistors that behave in a similar way to the synapses of neurons within the human brain.
A single neuron within the human brain can easily perform complex nonlinear calculations that are unmatched by logic functions in computer architectures. Researchers across the globe are trying to develop electronics that mimic how the brain computes to build high-performance energy-efficient devices. Although processors now have gotten faster over time, very few of them can compete with the speed and computing power of the human brain. Moreover, none of the existing processors come close to the brain's energy efficiency.
Memristors or memory resistors are passive devices that carry a memory of their past. When a voltage applied across the memristor is turned off, it remembers how much voltage was applied and for how long. These devices require much less power than transistors as they combine the functions of data storage and data processing in one unit, and data doesn't have to be transferred between different units.
Enter Lithionics is a term coined by Rupp and her students for the new field. "The vision of lithionics is, what other functional devices can we create with lithium that go beyond batteries to store, transfer and compute information?" says Rupp.
Rupp and Dr. Saeed Bastani of Ericsson are principal investigators for a team of four research groups working on applying lithium-based materials to next-generation neuromorphic systems (systems that mimic the structure of the human brain).
One of the research teams will use computational models to predict the best lithium composition for computing applications. This work will guide another team to make new materials and integrate them into prototype memristor units.
A third team will work on chip integration technologies. Finally, the fourth team will explore computer architectures that can efficiently utilize new devices. The researchers from Ericsson will participate in the evaluation of algorithms and hardware architectures of lithionic computing devices.
According to Rupp, such interdisciplinary nature of the collaboration is key. In a review of lithionics published in a 2020 issue of Nature Reviews Materials, Rupp and the team wrote that such collaborations "will lead to new material chemistries and device structures to achieve the ambitious goal of using lithium ions to power, compute and sense the world."
With the increasing demands of highly integrated sensor-based ecosystems for various applications, the number of devices connected to the internet is expected to increase dramatically. The second project of the MIT and Ericsson partnership aims to create truly wireless and autonomous electronic devices that do not need batteries or external power. The devices will be powered directly through the radio frequency (RF) signals abundantly present in the ambient environment.
Energy harvesting systems such as RF energy harvesters have emerged as a result of the emergence of ultra-low-power devices. RF harvesters are convenient for wireless systems as RF harvester circuits can be easily fabricated in such systems.
Image courtesy of Ericssson.
Jonas Hansryd, Manager of Microwave Systems Research at Ericsson Research and the company's project lead said, "This doesn't mean that all devices will be powered like this, but to have some devices that can be powered through the network would be really revolutionary and disruptive."
As a part of this project, researchers will work on RF energy harvester signals and build circuits that can operate with the extremely low levels of energy provided by RF signals.
Researchers believe that their innovations could one day fully enable the Internet of Things, or the increasingly connected world expected with next-generation mobile cellular networks, from 5G to the coming 6G and beyond.