The microprocessors and other chips that make our modern, digital lives possible are fabricated from wafers of highly refined silicon. This semiconducting element is ideally suited for the role — it is abundant and inexpensive, stable, and relatively easy to fabricate. Moreover, electrons can rapidly zoom through the features that are etched into the surface of the silicon that give it its function, which means very high speed components can be produced with it. When you have got a good thing going, why mess with it? That has more or less been the attitude of most people for several decades now. But as it turns out, we are rapidly approaching the physical limits of what is possible with silicon, and as we come closer to that limit, circuit behavior will become unstable and difficult to control.
The implication of this dilemma is that we need to start looking for alternatives now, before we are hamstrung by a substrate that is inadequate to meet our needs. A team at Harvard University has been exploring an alternative that might seem like a highly unlikely substitute at first glance. They recently published their work on an ionic computing chip, in which ions in an aqueous solution are manipulated to process information. Since ionic circuits — at least as they are presently implemented — cannot be as fast or accurate as their silicon counterparts, this may not seem like a very promising area of research. But what ionic circuits are lacking in speed and precision, researchers are hoping they can make up for with the diversity of ionic species that can be employed. Each species has unique physical and chemical properties that could potentially be leveraged to create richer information processing devices.
Previously, only individual electronic components, such as transistors and diodes, had been developed using ionic computing techniques. The Harvard team has moved the field forward by creating a single ionic computing chip containing hundreds of ionic transistors. This does not exactly have silicon shaking in its boots — silicon-based chips had achieved this feat by the 1960s, and modern chips have tens of billions of transistors, but it is an important step forward in the relatively new field of ionic computing, nonetheless.
The chip was designed for matrix multiplications (📷: W. Jung/Harvard SEAS)
The specific technique used by the team to create aqueous transistors involves a series of ring-shaped electrodes organized like a bullseye. These electrodes are surrounded by quinone molecules, and the outer electrodes electrochemically alter the pH level around the center electrode. The center electrode generates an electric current in the surrounding water, and that current level can be controlled by changing the pH level. Accordingly, the outer electrodes act as a sort of controllable gate that limits current flowing from the center electrode — and that, folks, is how you make a transistor with water.
Computing in an aqueous environment is much more reminiscent of brain activity than it is of traditional, silicon-based computation, so the team saw fit to develop an artificial neural network (or at least an important part of one) with their new chip. They created a 16×16 array of transistors to allow for analog matrix multiplication calculations. The current produced by the center electrode of each transistor is an arithmetic multiplication of the electrode’s voltage, and the pH level serves as a weight parameter. This type of calculation is the most prevalent one performed in neural networks today.
This work is still in the very early stages, and you will not find an ionic chip in your smartphone any time soon. However, as more ionic species are incorporated into these types of chips, and fabrication techniques improve, it will be interesting to see where this technology leads us in the future.