A programmable chemical computer with memory and pattern recognition

Area of Science:

• Chemical Engineering
• Computational Chemistry
• Biochemistry

Background:

• The von Neumann bottleneck limits current computer performance due to restricted data transfer between processing and memory units.
• Chemical reactions offer a potential alternative computing paradigm where processing and memory coexist, but programmability has been a major challenge.

Purpose of the Study:

• To develop a programmable chemical processor that overcomes the limitations of traditional computing architectures.
• To demonstrate the feasibility of using chemical reactions for data storage and pattern recognition.

Main Methods:

• A 5x5 array of cells utilizing the switchable oscillating Belousov-Zhabotinsky (BZ) reaction was developed.
• Individual cells were programmed to an 'on' or 'off' state, with distinct oscillation amplitudes detected via image processing.
• Interconnected BZ reactions were programmed to perform specific computational tasks.

Main Results:

• The system achieved over 2.9 x 10^17 possible chemical states, enabling high-density data representation.
• Chemically encoded and addressable memory was successfully demonstrated.
• A chemical Autoencoder was created, capable of pattern recognition at a rate equivalent to one million operations per second.

Conclusions:

• Programmable chemical processors based on the Belousov-Zhabotinsky reaction offer a viable path beyond the von Neumann bottleneck.
• This technology enables novel approaches to data storage and complex computational tasks like pattern recognition using chemical systems.