Researchers at Harvard Developed an Ionic Circuit Comprising Hundreds of Ionic Transistors and Performed a Core Process of Neural Network Computing in Water

Information is processed by microprocessors in computers, data centers, and smartphones by manipulating electrons that pass through solid semiconductors. However, our brains have a distinct mechanism as they process information by controlling ions in a liquid medium. For a very long time, scientists have worked to create “ionics” in an aqueous solution that mimics how the human brain processes information. Scientists think the variety of ionic species with different physical and chemical properties could be utilized for richer and more diverse information processing, despite the fact that ions in water move more slowly than electrons in semiconductors.

Ionic transistors and diodes have only been created as individual components in laboratories; no one has ever been able to combine numerous components to form a sophisticated circuit. The study of ionic computing is still in its early phases, though. Researchers from the Harvard John A. Paulson School of Engineering and Applied Sciences (SEAS) and DNA Script, a biotech startup, have created an ionic circuit with hundreds of ionic transistors and carried out a fundamental neural network computing operation as a first step towards producing such ground-breaking research. The study was also published in the Advanced Materials journal, and the chips mentioned are still in the prototype stage.

The researchers recently invented a method that was the foundation for the novel ionic transistor. The transistor is made of an aqueous solution of quinone molecules connected to two concentric ring electrodes and a bullseye-shaped center disc electrode. The two ring electrodes electrochemically reduce and regulate the local pH around the center disc by creating and trapping hydrogen ions. An ionic current flows from the center disc into the water due to an electrochemical reaction when voltage is applied. By adjusting the local pH, the reaction rate can be sped up or down, raising or decreasing the ionic current.

The next stage of their research involved designing the pH-gated ionic transistor so that the disc current results from adding the disc voltage and a “weight” parameter in arithmetic. This weight parameter represents the local pH gating of the transistor. The array of local pH values served as the weight matrix seen in neural networks. The transistors were placed into a 16 x 16 array to expand the analog arithmetic multiplication of individual transistors into an analog matrix multiplication.

Matrix multiplication, the most common calculation in neural networks for artificial intelligence, was used to analyze the ionic circuits. Based on electrochemical machinery, the team’s ionic circuit executes matrix multiplication in water in an analog fashion. To conduct matrix multiplication, microprocessors digitally manipulate electrons. The researchers emphasize that while the electrochemical matrix multiplication in water cannot be as quick or precise as digital microprocessors, it is attractive in its own right and has the potential to be energy-efficient.

The ionic circuit also has the ability to speed up processes like DNA synthesis and others involved in brain networks. Only a few ionic species, including hydrogen and quinone ions, have now been examined. However, as more ionic species are tried throughout time, information processing will only get more prosperous and more diverse. The team speculates that neural networks could soon operate on water-based ionic circuits, which would be substantially slower but far more energy-efficient.

References:

• https://seas.harvard.edu/news/2022/09/neural-net-computing-water
• https://www.pcgamer.com/water-chips-ionics-harvard-research/

Please Don't Forget To Join Our ML Subreddit