The article, titled “Chemical reservoir computation in a self-organizing reaction network” and published on 26 June 2024 in nature, discusses how formose reaction—a self-organizing network that turns formaldehyde into sugars—is used as a chemical “reservoir computer.” In a continuous stirred tank reactor, the Huck Research Group from the Radboud Universiteit (SRU) fed controlled inputs (formaldehyde, dihydroxyacetone, NaOH, CaCl₂), then monitored hundreds of chemical species in real-time using ion-mobility mass spectrometry.

Key highlights of the paper include:

• Self organized formose reaction as a chemical “reservoir computer”
  The team used the complex, self-organizing formose reaction in a continuous stirred tank reactor, feeding in controlled concentrations of formaldehyde, dihydroxyacetone, NaOH, and CaCl₂. The resulting mixture of ~106 distinct ions (measured via ion mobility mass spectrometry) forms a high dimensional dynamic system whose outputs, once linearly read out, perform computation.
• Wide range of nonlinear tasks without redesign
  This chemical reservoir excels in parallel computations: it can emulate all Boolean logic gates as well as more complex nonlinear classification tasks (e.g., XOR, concentric circles, sine patterns). On benchmark accuracy metrics, it matches or surpasses standard machine learning models—despite only using a fixed chemical network and adjustable read out weights.
• Modelling of dynamics and forecasting of chaotic systems
  Beyond static classification, the system was trained to predict the behavior of complex biochemical networks (like a simulated E. coli metabolic model) and to forecast future states of chaotic inputs (a Lorenz attractor). Crucially, certain chemical species in the reservoir act as short term memory nodes, retaining past environmental signals to enable accurate forecasting.

These findings provide compelling evidence that self organized chemical reaction networks can emerge as powerful and flexible information processors—offering a biomimetic route toward scalable, non digital computing systems.

Watch the video below to learn more.