Power/Performance Bits: May 19

Neuromorphic magnetic nanowires
Researchers from the University of Texas at Austin, University of Texas at Dallas, and Sandia National Laboratory propose a neuromorphic computing method using magnetic components. The team says this approach can cut the energy cost of training neural networks.

“Right now, the methods for training your neural networks are very energy-intensive,” said Jean Anne Incorvia, an assistant professor in the UT Austin Cockrell School’s Department of Electrical and Computer Engineering. “What our work can do is help reduce the training effort and energy costs.”

The team’s system uses magnetic nanowires (the artificial neurons) spaced in particular ways that increase the ability for the artificial neurons to compete against each other, with the most activated ones winning out. This effect, called lateral inhibition, traditionally requires extra circuitry, increasing costs and requiring more energy and space.

Lateral inhibition occurs when the brain’s neurons firing the fastest are able to prevent slower neurons from firing. In computing, this cuts down on energy use in processing data.

Incorvia said their method provides an energy reduction of 20 to 30 times the amount used by a standard back-propagation algorithm when performing the same learning tasks.

This research focused on interactions between two magnetic neurons and initial results on interactions of multiple neurons. Next, the team will work toward applying the findings to larger sets of multiple neurons as well as experimental verification of the findings.

Two record-breaking solar cells
Two groups of researchers announced record-breaking solar cells in different categories.

In one, scientists at the National Renewable Laboratory (NREL) fabricated a six-junction solar cell with the world’s highest solar conversion efficiency at 47.1%, which was measured under concentrated illumination. A variation of the same cell also set the efficiency record under one-sun illumination at 39.2%.

“This device really demonstrates the extraordinary potential of multijunction solar cells,” said John Geisz, a principal scientist in the High-Efficiency Crystalline Photovoltaics Group at NREL.

Multi-junction solar uses different materials to absorb different wavelengths of light. The device contains about 140 total layers of various III-V materials to support the performance of these junctions, and yet is three times narrower than a human hair. Due to their highly efficient nature and the cost associated with making them, III-V solar cells are most often used to power satellites. Other applications include solar concentrator system.

Geisz said that currently the main research hurdle to topping 50% efficiency is to reduce the resistive barriers inside the cell that impede the flow of current. NREL is also engaged in reducing the cost of III-V solar cells.

For the other, researchers at University of Cambridge, Helmholtz-Zentrum Berlin (HZB), Eindhoven University of Technology, and Salerno University built a perovskite CIGS tandem solar cell that achieved an efficiency of 24.16%.

Like the previous cell, this one uses multiple materials to capture more of the light spectrum. Metal-halide perovskite compounds mainly use the visible parts of the spectrum, while CIGS semiconductors convert rather the infrared light.

CIGS cells, which consist of copper, indium, gallium and selenium, can be deposited as thin-films with a total thickness of only 3 to 4 micrometers; the perovskite layers are even much thinner at 0.5 micrometers. The new tandem solar cell made of CIGS and perovskite thus has a thickness of well below 5 micrometers, which would allow the production of flexible solar modules.

“This combination is also extremely light weight and stable against irradiation, and could be suitable for applications in satellite technology in space,” said Prof. Dr. Steve Albrecht of HZB.

Small, fast modulator
Researchers at George Washington University, University of Texas at Austin, and Omega Optics developed a silicon-based electro-optical modulator they say is smaller, as fast as, and more efficient than state-of-the-art technologies.

Used for converting electrical data to optical data streams, typical electro-optical modulators are between 1mm and 1cm. The team’s new device is 1um, and capable of providing gigahertz-fast signal modulation.

To make the device, the team combined indium tin oxide (ITO), the transparent, conductive material often used in touchscreen displays and solar cells, with a silicon photonic chip platform.

While silicon often serves as the passive structure on which photonic integrated circuits are built, the light matter interaction of silicon materials induces a rather weak optical index change, requiring a larger device footprint. The researchers found that adding a thin ITO layer to the silicon photonic waveguide provided an optical index change 1,000 times larger than silicon.

According to the team, the device is stable against temperature changes and allows a single fiber-optic cable to carry multiple wavelengths of light, increasing the amount of data that can move through a system.