Research Bits: Oct. 25

Polarization for photonic processor

Researchers from the University of Oxford and University of Exeter developed a photonic processor that uses multiple polarization channels, increasing information density.

“We all know that the advantage of photonics over electronics is that light is faster and more functional over large bandwidths. So, our aim was to fully harness such advantages of photonics combining with tunable material to realize faster and denser information processing,” said June Sang Lee, a DPhil student in the Department of Materials, University of Oxford.

The team developed a HAD (hybridized-active-dielectric) nanowire, using a hybrid glassy material which shows switchable material properties upon the illumination of optical pulses. Each nanowire shows selective responses to a specific polarization direction, so information can be simultaneously processed using multiple polarizations in different directions.

Using this concept, the researchers designed a photonic processor in which computation is carried out through multiple polarization channels. The nanowires are modulated by nanosecond optical pulses, increasing computing speed. According to the researchers, the new chip could be more than 300 times faster and denser than current chips.

“This is just the beginning of what we would like to see in future, which is the exploitation of all degrees of freedoms that light offers, including polarization to dramatically parallelize information processing,” said Harish Bhaskaran, a professor in the Department of Materials, University of Oxford. “Definitely early-stage work – our speed estimates still need research to verify them experimentally – but super exciting ideas that combine electronics, non-linear materials and computing.”

FeFETs for physical reservoir computing

Engineers at the University of Tokyo created a physical reservoir computing system using hafnium-oxide ferroelectric materials.

Physical reservoir computing (PRC) is a way of creating neural networks in hardware. It is based on recurrent neural networks (RNNs), which process data over time rather than static data, taking into account information from previous inputs to consider a current input and the output. RNNs are suitable for applications such as speech recognition, natural language processing, or language translation  where the data’s sequence or time-based context is important.

“In physical reservoir computing, the input data are mapped onto patterns in some physical system, or reservoir (such as the patterns in the structure of a magnetic material, a system of photons, or a mechanical device), that enjoys a higher dimensional space than the input,” the researchers explain. “Then, a pattern analysis is performed on spatio-temporal patterns on the final readout “layer” to understand the state of the reservoir. Because the AI is not trained on the recurrent connections within the reservoir, but only on the readout, simpler learning algorithms are achievable, dramatically reducing the computation required, enabling high-speed learning and lowering the energy consumption.”

The team’s PRC architecture uses ferroelectric gate transistors (FeFETs) made of hafnium oxide-based ferroelectric materials. “These materials are already commonly used in semiconductor integrated circuit manufacturing processes,” said Shinichi Takagi, a professor with the Department of Electrical Engineering and Information Systems at the University of Tokyo. “This means that FeFET reservoirs are expected to be integrated with large-scale semiconductor integrated circuit fabrication with little difficulty compared to some more novel material.”

In testing on a speech recognition application, the team found the PRC architecture to be 95.9% accurate for speech recognition of the numbers zero to nine. They plan to see if they can increase the computing performance and test the FeFET reservoirs on other applications.

5G and wireless power

Scientists from the Tokyo Institute of Technology built a wirelessly powered transmitter-receiver for 5G networks.

“The millimeter-wave wireless power transfer system is a promising solution for massive Internet of Things, yet it has been hampered by technical problems. We were thus able to make a breakthrough by producing a 5G transceiver with high efficiency at big angles and distances,” said Atsushi Shirane, an associate professor at Tokyo Tech.

The device has two modes, a receiving mode and a transmitting mode. In the receiving mode, the device receives a 5G signal and a millimeter-wave power signal. This power signal activates the device and provides it with power. The device then enters the transmission mode and sends a 5G signal back in the same direction from which it initially received one. This would enable an IoT device to communicate without needing to be plugged in. The researchers note that the device can generate power over a wide span of angles and distances, a challenge for previous wireless power devices.

“This was the world’s first simultaneous reception of power and communication signals with beam steering,” said Shirane. “We truly believe that technology like this can revolutionize the Internet of Things network and free it from the shackles that bind it today.”