Power/Performance Bits: Aug. 21

Physical neural network
Engineers at UCLA built a physical artificial neural network capable of identifying objects as light passes through a series of 3D printed polymer layers.

Called a “diffractive deep neural network,” it uses the light bouncing from the object itself to identify that object, a process that consumes no energy and is faster than traditional computer-based methods of image identification.

The process of creating the artificial neural network began with a computer-simulated design. Then, the researchers used a 3D printer to create very thin, 8 centimeter-square polymer wafers. Each wafer has uneven surfaces, which help diffract light coming from the object in different directions. The layers look opaque to the eye but submillimeter-wavelength terahertz frequencies of light used in the experiments can travel through them. And each layer is composed of tens of thousands of artificial neurons — in this case, tiny pixels that the light travels through.

Together, a series of pixelated layers functions as an “optical network” that shapes how incoming light from the object travels through them. The network identifies an object because the light coming from the object is mostly diffracted toward a single pixel that is assigned to that type of object.

The researchers then trained the network using a computer to identify the objects in front of it by learning the pattern of diffracted light each object produces as the light from that object passes through the device.

The network, composed of a series of polymer layers, works using light that travels through it. Each layer is 8 centimeters square. (Source: UCLA Samueli / Ozcan Research Group)

“This is intuitively like a very complex maze of glass and mirrors,” said Aydogan Ozcan, UCLA Chancellor’s Professor of Electrical and Computer Engineering. “The light enters a diffractive network and bounces around the maze until it exits. The system determines what the object is by where most of the light ends up exiting.”

The device could accurately identify handwritten numbers and items of clothing, two commonly used tests in artificial intelligence studies, in experiments where the researchers placed images in front of a terahertz light source and let the device “see” those images through optical diffraction.

They also trained the device to act as a lens that projects the image of an object placed in front of the optical network to the other side of it.

Because its components can be created by a 3D printer, the artificial neural network can be made with larger and additional layers, the team said, resulting in a device with hundreds of millions of artificial neurons. Those bigger devices could identify many more objects at the same time or perform more complex data analysis. The components can be made inexpensively — the device created by the UCLA team could be reproduced for less than $50.

While the study used light in the terahertz frequencies, Ozcan said it would also be possible to create neural networks that use visible, infrared or other frequencies of light. A network could also be made using lithography or other printing techniques, he said.

“This optical artificial neural network device is intuitively modeled on how the brain processes information,” said Ozcan. “It could be scaled up to enable new camera designs and unique optical components that work passively in medical technologies, robotics, security or any application where image and video data are essential.”

Recycling Li-ion batteries
Researchers at Michigan Technological University adapted 20^th century mining technologies used to separate metal from ore to improve lithium-ion battery recycling.

The team used mining industry technologies to separate everything in the battery: the casing, metal foils and coatings for the anode and cathode, which includes lithium metal oxide, the most valuable part. The components can be returned to the manufacturer and re-made into new batteries.

The process is also inexpensive and energy efficient.

After trying a range of solvents to liberate the different chemicals, the team turned to water and kerosene.

“We use standard gravity separations to separate copper from aluminum, and we use froth flotation to recover critical materials, including graphite, lithium and cobalt. These mining technologies are the cheapest available, and the infrastructure to implement them already exists,” said Lei Pan, an assistant professor of chemical engineering at Michigan Technological University.

Froth floatation separates hydrophobic and hydrophilic materials. The process involves crushing or grinding a combined material, which is added to water to form a slurry. A collector chemical, in this case kerosene, is added to make the desired material hydrophobic. The slurry is aerated, producing air bubbles to which the hydrophobic material attaches as they rise to the top, forming a froth. The materials remaining in the slurry are referred to as tails or tailings.

In their experiments, the team found that over 90% of anode materials were floated in froth layers, while 10–30% of cathode materials were floated.

The team sees ways to further improve the purity of the separated materials. “For spent lithium-ion batteries, a low purity of cathode materials in tailings might be improved by fine grinding, at which freshly liberated hydrophobic surfaces are exposed and consequently anode materials become floatable,” they said. “The present result confirms that the froth flotation technique is a viable and versatile technique in producing high purity cathode materials from lithium-ion batteries.”

Solar supercapacitor for flexible sensors
Researchers from the University of Glasgow developed a promising new type of graphene supercapacitor, which they think shows promise for use in the next generation of wearable health sensors.

The new supercapacitor uses layers of flexible, three-dimensional porous foam formed from graphene and silver to produce a device capable of storing and releasing around three times more power than any similar flexible supercapacitor.

In demonstrations, the supercapacitor was capable of providing power across 25,000 charging and discharging cycles with 68% capacitance retention. Additionally, the observed energy and power densities were found to be better than the values reported for carbon-based supercapacitors.

The team integrated this supercapacitor with a flexible solar cell ‘skin’, one of the group’s previous developments, effectively creating an entirely self-charging system that was used to power a pH sensor which uses a wearer’s sweat to monitor their health.

An illustration of the solar-powered wearable pH sensor. (Source: Libu Manjakkal, et al. / University of Glasgow / Nano Energy Volume 51, September 2018)

“We’re very pleased by the progress this new form of solar-powered supercapacitor represents. A flexible, wearable health monitoring system which only requires exposure to sunlight to charge has a lot of obvious commercial appeal, but the underlying technology has a great deal of additional potential,” said Ravinder Dahiya, a professor at the University of Glasgow.

“This research could take the wearable systems for health monitoring to remote parts of the world where solar power is often the most reliable source of energy, and it could also increase the efficiency of hybrid electric vehicles. We’re already looking at further integrating the technology into flexible synthetic skin which we’re developing for use in advanced prosthetics.”