A laboratory proof of concept used a self-promoting chemical reaction to classify simple binary images. Rather than calculating with electronic components, the system used reaction timing as the nonlinear mechanism in a winner-take-all network: the candidate class whose reaction reached a transition first was selected.
How can a chemical reaction work like a neural network?
A neural network combines inputs according to weights and uses nonlinear behavior to distinguish among possible outputs. In this experiment, the weighted inputs were represented by chemical samples, and reaction kinetics supplied the nonlinear behavior. The researchers used a winner-take-all design: candidate classes raced, and the first reaction to meet a specified condition determined the output.
As an Amazon Associate I earn from qualifying purchases.
The chemical process was a copper-catalysed click reaction between 2-azidoethanol and tripropargylamine. It produces tris(triazolylmethyl)amine, a product that increases the production and activity of copper(I), or Cu(I). Because the reaction product helps promote further reaction activity, the process is autocatalytic.
Recommended Free Tools
The team used the reaction’s transition time, denoted t1/2, as the selection condition. The initial concentration of tris(triazolylmethyl)amine could tune that time. During training, the network’s weights were adjusted so that the class most similar to the input would be the one whose reaction reached the transition first. Chemistry World described the click reaction as “really robust,” quoting infochemistry researcher Ekaterina Skorb of ITMO University; that comment concerns the reaction, not the readiness of the computing system for practical deployment.
#1 Best Overall
How were images encoded and classified?
The demonstration used binary images measuring 16×16 pixels. Each pixel corresponded to a position on a well plate, and its color was represented by the initial catalyst concentration at that position.
- Encode the image: Set the catalyst concentration in each well to represent the corresponding binary pixel.
- Apply the weights: An automated liquid handler removed weight-dependent volumes from the wells and pooled samples for each candidate image class.
- Start the candidate reactions: Fresh reagents were added to the pools.
- Identify the winner: The pools were monitored with UV-Vis spectroscopy; the candidate whose reaction reached the transition first was selected as the classification.
In other words, the image was not scanned by a camera and classified by a conventional digital model. Its encoded chemical concentrations were physically combined according to the network’s weights, then the reaction timing distinguished among candidate classes.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What did the experiment demonstrate—and what did it not?
The demonstration showed that reaction kinetics can serve as the nonlinear element of a winner-take-all classifier, with automated liquid handling connecting a pixel-based input to candidate reactions. Chemistry World’s report identifies the underlying paper as C. Arcadia et al., published in Chemical Science in 2021 (DOI: 10.1039/d0sc05860b).
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The reported account describes classification of encoded binary images, but it gives no named accuracy figure, sample size, or comparative benchmark. It therefore does not establish that the approach is more accurate, faster, cheaper, or more energy-efficient than electronic neural networks. Nor does it demonstrate a general-purpose computer, broad scalability, or a system ready for practical deployment.
Rank #3
The contrast with electronic classifiers is primarily one of substrate and mechanism: this proof of concept represented inputs as catalyst concentrations and used chemical reaction timing, with robotic handling and spectroscopy, rather than electronic computation. Without comparable measurements, the experiment cannot establish which approach performs better for image classification.
Quick Recap
Best Value
Rank #4
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
