When it comes to computing, humanity has followed a predominantly digital path and has become very proficient in this field. But computing goes beyond that and should be generally applicable to both living and inert systems, something Alan Turing himself believed. ‘In my opinion, the true nature of computing can be derived from how living systems communicate and compute with chemical inputs, using logic’, says Associate Professor Albert Wong of the University of Twente. His research group studies chemical reaction networks. 

Building block

Wong explains that as far as their development is concerned, chemical computing systems may never compete with the speed and flexibility of digital computing. ‘At the same time, there is evidence that the current way we store and process information is simply unsustainable, particularly with the rise of AI. So our goal is not to improve digital computing, but to have a radically different approach. And now we may have created the first building block for a real chemical computer.’

Wong, Dmitrii Kriukov and Jurriaan Huskens developed a strategy that uses metal ions to control the rate of a trypsin-catalysed autocatalytic reaction network. ‘At the beginning of this project, I asked Dmitrii a deliberately open question: ‘How can we use molecules for computation?’. Over the next two and a half years of his PhD, Kriukov came up with a plan and conducted experiments and simulations all on his own, resulting in a fascinating alternative to conventional computing. 

Neurons

The idea came from his ‘basic and shallow’ ideas of decision-making in neurons, says Kriukov. ‘When making decision-like actions, neurons either fire or stay suppressed depending on nonlinear integration of information that comes from dendrites. When a neuron is triggered by an external chemical signal, it can exhibit long-term potentiation and long-term depression of a signal transmission, so, behave dependent on the context of coming information. I wanted to do something similar with molecules.’

In order to get chemical systems into a state where they can respond to incoming information, you have to meet a few requirements, Kriukov explains. ‘The most important is that the reaction should be highly non-linear and have intrinsic feedback, and the most fundamental example of such behaviour is the autocatalytic reaction.’ The team created two non-linear ‘layers’ in the system: first, history dependence, that comes from intrinsic feedback of autocatalysis, and second, Boolean logic, that comes from nonlinear control over the rate of the autocatalysis. ‘These two things combined enable the system to remember its nonlinear decisions.’

Put simply, you have a chemical system that can be used as a mathematical operation with a ‘save’ function. ‘This is very new in chemical systems’, Kriukov continues. ‘Normally, if you want to go from one Boolean task to another, you have to change a lot of things, sometimes even the entire molecular system. But we have a continuum of mathematical functions, and we are the first to show a variety of logic operations within a narrow window of parameters.’

Blueprint

They kept the trypsin-catalysed autocatalytic reaction network out of equilibrium in a flow setup and were able to control the rate of the reaction with the metal ions Ca²⁺, La³⁺ and Nd³⁺. Kriukov: ‘Depending on the concentration, neodymium can either accelerate or inhibit the reaction, while calcium and lanthanum both accelerate it, but in different ways. This makes the system even more flexible and gave us the ability to set up mathematical functions such as not-OR, exclude-OR and not-AND, which are considered very difficult to for chemical systems.’

‘What I find so impressive about Dmitrii’s idea is that it’s simple enough to work’, says Wong. ‘You don’t need complicated devices, which is important to keep the whole neuromorphic computer simple. We’re not there yet, but what we have so far is the beginning of a blueprint for a chemical computer that has never been seen before’.

In their paper, the authors write that this work is likely to have an impact on many scientific fields. ‘Dmitrii and I are both associated with the BRAINS centre for brain-inspired nanosystems at Twente’, says Wong. ‘We see people using photonics, electronics and other systems to mimic the human brain. We hope that our chemical building block will inspire others to think differently about computing and take development of future technology to a new level. Who knows what the computer will look like in twenty years’ time?’

Wong, A., Kriukov, D. and Huskens, J. (2024) Nat. Commun. 15, DOI: 10.1038/s41467-024-52649-z