Power/Performance Bits: July 23

Image-recognizing glass
Engineers at the University of Wisconsin-Madison, MIT, and Columbia University developed a way to create ‘smart’ glass capable of performing image recognition tasks without the need for electronics or power.

“We’re using optics to condense the normal setup of cameras, sensors and deep neural networks into a single piece of thin glass,” said Zongfu Yu, electrical and computer engineering professor at UW-Madison. “This is completely different from the typical route to machine vision.”

The system uses translucent squares of glass with tiny bubbles and impurities embedded within, which bend light in specific ways to differentiate images.

As a proof of concept, the researchers created glass pieces to identify handwritten numbers. Designing the glass was similar to a machine learning training process, the team said, except with an analog material. Along with air bubbles of different sizes and shapres, small pieces of light-absorbing materials like graphene were placed at specific locations.

Light emanating from an image of a number enters at one end of the glass, and then focuses to one of nine specific spots on the other side, each corresponding to individual digits. The device was able to detect when a handwritten 3 was changed to an 8 in real-time.

“We’re accustomed to digital computing, but this has broadened our view,” said Yu. “The wave dynamics of light propagation provide a new way to perform analog artificial neural computing.”

In particular, once the material is designed, the computation is passive and instrinsic. A single piece tuned to a specific task could be used many times. “We could potentially use the glass as a biometric lock, tuned to recognize only one person’s face,” said Yu. “Once built, it would last forever without needing power or internet, meaning it could keep something safe for you even after thousands of years.”

“The true power of this technology lies in its ability to handle much more complex classification tasks instantly without any energy consumption,” said Ming Yuan, professor of statistics at Columbia University. “These tasks are the key to create artificial intelligence: to teach driverless cars to recognize a traffic signal, to enable voice control in consumer devices, among numerous other examples.”

While the initial training process could be time consuming and computationally demanding, the team says that fabricating the glass itself is easy and inexpensive. In the future, the researchers plan to determine if their approach works for more complex tasks, such as facial recognition.

Metabolite storage
Researchers at Brown University developed a molecular way to store data using arrays of metabolites. These liquid mixtures containing sugars, amino acids, and other types of small molecules make up metabolomes capable of encoding kilobyte-scale image files that could then be read back.

Molecular data storage is an area seeing much investigation, though most of it focuses on using DNA molecules. “This is a proof-of-concept that we hope makes people think about using wider ranges of molecules to store information,” said Jacob Rosenstein, a professor in Brown’s School of Engineering. “In some situations, small molecules like the ones we used here can have even greater information density than DNA.”

A metabolome is the collection of molecules an organism uses to regulate metabolism. The researchers created their own artificial metabolomes, a liquid mixure with different combinations of molecules in which the presence or absence of a particular metabolite encodes one bit of digital data. The number of molecule types in the artificial metabolome determines the number of bits each mixture can hold, and for this study the team created libraries containing six or twelve metabolites, such that each mixture could encode six or twelve bits.

Thousands of these mixtures were arrayed in nanoliter-sized droplets on small metal plates, which were then dried, leaving the metabolite molecules. The data was read back using a mass spectrometer. The researchers used the technique to encode and retrieve a variety of image files of sizes up to 2 kilobytes.

The researchers note that while the capacity so far is small, there is potential to scale up. One limitation is that many metabolites chemically interact with each other when placed in the same solution, and that could result in errors or loss of data.

However, Rosenstein argues that this creates the potential for molecular systems that not only store data, but could also perform computations within metabolite mixtures. “Research like this challenges what people see as being possible in molecular data systems,” Rosenstein said. “DNA is not the only molecule that can be used to store and process information. It’s exciting to recognize that there are other possibilities out there with great potential.”

Solar window film
Researchers at Chalmers University of Technology, University of Copenhagen, and University of Adelaide developed a film capable of absorbing solar heat during the day and slowly releasing it for nighttime heating.

The film, which the researchers envision as a window laminate, utilizes norbornadiene (NBD)–quadricyclane (QC) molecular photoswitches embedded into polymer matrices. When these photoswitches are struck by solar radiation, they capture photons and isomerise, changing form.

When the film has not absorbed any solar energy, it is yellow or orange. When the molecule captures solar energy and is isomerised, it loses its color and then becomes entirely transparent. As long as the sun is shining on the film it captures energy, so less heat penetrates through the film and into the room.

At dusk, when there is less sunlight, heat starts to be released from the film and it gradually returns to its yellow shade and is ready to capture sunlight again the following day.

“For example, airports and office complexes should be able to reduce their energy consumption while also creating a more pleasant climate with our film, since the current heating and cooling systems often do not keep up with rapid temperature fluctuations,” said Kasper Moth-Poulsen, a chemist at Chalmers.

The film’s creation is part of a larger project on the concept of Molecular Solar Thermal Storage. Previously, the team created a system that was capable of storing solar energy for extended periods, but think the film is a closer step to application. The researchers still need to increase the concentration of the molecule in the film and bring down the price for it to be commercially viable, but say they are close.