Research Bits: June 5

Improving memristors
Researchers at Los Alamos National Laboratory (LANL) have demonstrated a reliable Interface-type (IT) memristive device (memristor) that shows promise as a technique for building artificial synapses in neuromorphic computing.

The team made its memristor — a component that which combines memory and programming functions — using a simple Au/Nb-doped SrTiO3 (Nb:STO) Schottky structure that overcomes some of the issues with IT memristors that use oxygen ions. The IT memristor had high repeatability, stability, and low device-to-device, cell-to-cell, and cycle-to-cycle variability during endurance and retention tests. An artificial neural network that the researchers simulated with Au/Nb:STO synapses was able to recognize images from the Modified National Standards and Technology database maintained by the National Institute of Standards and Technology (NIST) at an 94.7% accuracy rate.

The researchers’ paper is open access.

Kunwar, S., Jernigan, Z., Hughes, Z., Somodi, C., Saccone, M.D., Caravelli, F., Roy, P., Zhang, D., Wang, H., Jia, Q., MacManus-Driscoll, J.L., Kenyon, G., Sornborger, A., Nie, W. and Chen, A. (2023), An Interface-Type Memristive Device for Artificial Synapse and Neuromorphic Computing. Adv. Intell. Syst. 2300035. https://doi.org/10.1002/aisy.202300035

Organic electrochemical transistors
In an attempt to mimic the computing efficiency of biological nervous systems and cut down the noise in biological sensors, researchers have been working on organic electrochemical transistors (OECTs), which are transistors with a bulk field effect that can operate at low Hertz and in saline environments — which is what you find in the biological systems.

Researchers at Xi’an Jiaotong University in China, the University of Hong Kong, and Xi’an University of Science and Technology say they have designed an OECT that is capable of sensing, memory, and processing. (The co-location of the three functions is much more efficient than a silicon hardware in energy use and area.) Their OECT uses a vertical traverse architecture (v-OECT) and electrode process and can switch between the sensing and processing modes with doping a crystalline–amorphous channel with ions to achieve a volatile receptor or a non-volatile synapse. When a receptor, its sensing is multi-modal. The researchers say the OECTs can sense ion concentration changes in plants and record the electrocardiogram (ECG) signals, temperature sensations, gustation, and artificial vision. When the OECT is a non-volatile synapse, the researchers say it can “offer 1,024 (10-bit) distinct states, wide dynamic range and state retention of more than 10,000 s.”

Wang, S., Chen, X., Zhao, C. et al. An organic electrochemical transistor for multi-modal sensing, memory and processing. Nat Electron 6, 281–291 (2023). https://doi.org/10.1038/s41928-023-00950-y

For a good introduction to OECTs and why this type of transistor performs better in biological signal sensing, see this presentation from the 2019 ARM Research Summit by Cambridge University’s Chris Proctor.

High-speed thin-film lithium niobate quantum processor
Researchers at University of Copenhagen and University of Muenster have had success experimenting with a high-speed and reconfigurable quantum photonic processor using single-crystal thin films of lithium niobate — LiNbO3 (LN) — that are bonded on a silica-insulating substrate consisting of lithium niobate on insulator (LNOI). The reconfigurable LNOI quantum photonic processor controls the quantum states of light emitted from a quantum dot (QD) single-photon source (SPS).

Several qualities of the LN make it useful. LN has strong electro-optical properties, high transparency, high index contrast, and can handle cryogenic temperature.  The LN can keep the ICs smaller and able to handle gigahertz.  The researchers proposed an LNOI platform for photonic / quantum interface. Part of their experiment was to integrate fast phase shifters on the LNOI platform the quantum emitter wavelengths. The researchers generated photons that were processed with low-loss circuits programmable at speeds of several gigahertz and demonstrated an on-chip photon router for the QD-emitted photons.

Photonics used in quantum computing may help make quantum hardware scalable. Photonics are hoped to make long range quantum networks possible, serve as interconnects between more than one quantum device, and be used as large-scale circuitry for quantum computing and simulation.

Patrik I. Sund et al, High-speed thin-film lithium niobate quantum processor driven by a solid-state quantum emitter, Science Advances (2023). https://doi.org/10.1126/sciadv.adg7268