Soft Hardware, Flowing Software

This is a short pop-sci-ish article about my latest pre-print, check it out for details!

Chemical and physical computing is very cool, and also quite useful in the future. Especially in environments where traditional electronics don't work so well - say, inside living organisms, or in environments where electricity is not convenient - intelligence embedded into materials would be very nice to have. You would not need a computer to sense things, process the information, and act upon the inputs, but instead, have a material that does all of that in one go.

This field of chemical and physical computing is growing fast. Chemical computing is typically implemented using a chemical software, that contains molecules that react and do the job of the computation, much like software would in traditional computers. This software could for instance be a group of dissolved DNA molecules that together perform a specific action, like a logic AND gate.

The problem comes when combining too much of this chemical software. Chemistry tends to interact with other chemistry: there is a limit to the amount of unique DNA molecules you can put into something before it becomes a mess. Software can be combined in layers upon layers of abstraction, different software running on the same processor does not really matter. In physical systems, this remains difficult.

Our recent paper tries to alleviate a part of the problem. What if, instead of the chemistry determining everything, we make the surroundings of the chemistry important as well? I developed a reconfigurable microfluidic platform where soft gel structures can be printed and erased at will. Rather than etching permanent channels, the system sculpts temporary architectures that guide chemical reactions through specific pathways. Once the computation is complete, these structures dissolve away, leaving the space open for a completely different configuration. We design the chemistry - the molecular software - such that it can have multiple functions based on what is printed. The same DNA strands perform an AND operation in configuration 1, and an OR operation in configuration 2.

Finally, I used this reconfigurable microfluidics system as a simple physical reservoir computer. We put a complex chemical reaction inside the microfluidic chip. By printing different things inside the chip, the reaction gets pushed into a certain direction. The observation of what directions the reaction is nudged into, can then be used to do simple calculations. This works trough some surprisingly simple principles, but you should just read the paper to see that.

So, what have we learned? Using hardware (geometry, conditions) to modulate the software (chemistry) turns a simple microfluidic chip into a universal, reprogrammable, chemical computer—just print a new architecture to run a new program. Yes, things are still simple, don't expect me to run DOOM on this yet. But it genuinely brings us a step closer to intelligent materials that can operate inside living tissue or harsh environments, doing their thinking without silicon or copper wires.