Harvard-made Ionic Circuit Computes in Water
Researchers at Harvard claim to have developed the first ionic circuit based on a new ionic transistor technique.
In many respects, the human brain is the greatest computing system ever made. Meanwhile, the digital computers that today’s world runs on are a far cry away, leaving researchers to wonder where the discrepancy is and how we can bridge that gap.
One enormous difference between the human brain and a digital computer is how computation is done. In digital computers, electrons are manipulated through semiconductor materials. In the brain, ions are manipulated in liquid.
In an attempt to replicate this process, researchers at Harvard University recently published a paper describing a new ionic circuit. In this article, we’ll be talking about ionic computing as well as the new research from Harvard.
What is an Ionic Circuit?
By and large, the physiological processes that living organisms rely on are governed by the selective transport of ions across aqueous environments. For example, cells function based on the biological channels across their membranes which facilitate the exchange of ions and molecules both within the cell and external to the cell.
Example of an ionic bipolar junction transistor. Image used courtesy of from Lucas et al
To the discerning observer, one may realize that this, in itself, is a type of circuit, albeit not in the more conventional sense. And, while these “ionic circuits” are not necessarily conventional, it is well understood that, in many ways, their performance and characteristics far exceed those of digital computers.
Even though ions in water may move slower than electrons in semiconductors, many believe that the diversity of ionic species can lend themselves to a richer field of information processing that could benefit fields such as machine learning.
Hence, to try and harness this behavior, researchers have sought after ionic circuits, where the movement of ions in aqueous solutions can mimic conventional computing techniques. So far, research attempts have been successful in creating components for ionic circuitry, such as diodes or transistors, but no one has created a full circuit out of ionic components.
New Harvard Research
Researchers at Harvard University challenged that narrative when they published its paper describing a new ionic circuit for machine learning computation. Specifically, in their paper, the researchers describe creating an analog multiply-and-accumulate (MAC) block out of an ionic circuit.
The building block of the MAC was a novel type of ionic transistor that leveraged earlier research from Harvard in which researchers found that they were able to electronically program the pH of certain aqueous solutions.
The ionic chip from Harvard. Image used courtesy of Harvard University
In the MAC research, the team created an ionic transistor consisting of two concentric ring electrodes with a central disk electrode within a solution consisting of quinone molecules. The device works to electrochemically manipulate the pH around the center disk through the production of hydrogen atoms, at which point the center disk can generate an ionic current into the water that is proportional to the local pH. In this way, the pH of the solution acts as a transistor gate and effectively creates an ionic transistor.
Then, to go from transistor to MAC, the team architected the transistors in such a way that the total ionic current is equivalent to the multiplication of the individual transistor’s currents. These multiplication cells were then arranged in a 16x16 array to create a matrix multiplication block. Hence, using pH control and their unique ionic transistors, the researchers were able to make a MAC block with applications in machine learning computation.
While the Harvard research team admits that their new ionic MAC is nowhere near as fast as conventional digital MACs, they do believe it holds the potential to eventually be more efficient than conventional solutions. Ideally, the researchers hope that ionic circuits can serve as a more efficient and rich computing resource for applications like edge computing.