Researchers develop ionic chips for neural networks

Conventional chips based on semiconductors have long been anchored and established in everyday use. As researchers are just finding out, it doesn’t have to be just that, because ionic microprocessors are currently being developed, which are said to be particularly energy-efficient and are based on nature’s designs in their work. Instead of electricity, the processors, which are equipped with ionic transistors, will in future use electrochemistry to solve calculations.

A team from the John A. Paulson School of Engineering and Applied Sciences (SEAS) in Harvard, who have developed an ion circuit that includes hundreds of ion transistors, is currently researching this. This approach aims to utilize the processing systems of biology, with specific reference to the brain. According to research, ionic processors with their energy-efficient designs could be valuable for certain application scenarios in the future.

Electrochemistry as a solution for matrix multiplication

The researchers’ ionic circuit is thought to have been created by multiplying the number of individual ionic transistors that can work together. Their ionic transistor design consists of an aqueous solution of quinone molecules connected to two concentric ring electrodes (blue and red in the figure) and a third (yellow) disc electrode in the center, essentially forming a bullseye-like transistor design.






A CMOS chip (right) with an array (center) containing hundreds of individual ion transistors (right).

Source: Woo-Bin Jung/Harvard SEAS




As a consequence of voltage passed through the transistors, the concentric ring electrodes should be able to locally adjust the pH of the water. It does this by increasing and decreasing hydrogen ions. This creates an ionic current from the transistors that can act as an on and off switch, known as the gate. The transistors can process binary information by changing the pH value and the resulting ion current.

Also interesting: Less e-waste: Researchers demonstrate batteries made from paper, ink, salt and water

The analog transistors, which can represent the 0 and 1 states, were further set up by the researchers in a 16 by 16 matrix grid arrangement (grid array). This enables the ionic processor to also solve matrix multiplications. According to one of the researchers, such calculations are widely used in neural networks for artificial intelligence, which is why ionic processors could achieve a breakthrough in this field.

The research on the processors is not finished yet. Meanwhile, one would like to expand the electrochemical reactions for information processing. This includes, for example, increasing the number of ionic species that can be manipulated, since currently only three to four such species are used, which include hydrogen and quinone ions.

Source: Harvard University via Tom’s hardware

source site