‘Training’ carbon-nanotube composites in ‘unconventional’ computing

April 10, 2015

As the liquid crystals align in electric fields, it helps to align the nanotubes — changing the electrical structure of the materials. You can see the thermal output from the material during this “training” process. Bright colors represent localized heating within the material, which the group suspects is due to the formation of new conductive pathways as the material changes or evolves. (Credit: Mark K. Massey/Journal of Applied Physics)

Researchers from Durham University and the University of São Paulo-USP have  developed a method of using single-walled carbon nanotube (SWCNT) composites in “unconventional” computing.

By studying the mechanical and electrical properties of the materials, they discovered a correlation between SWCNT concentration/viscosity/conductivity and the computational capability of the composite.

“Instead of creating circuits from arrays of discrete components (transistors in digital electronics), our work takes a random disordered material and then ‘trains’ the material to produce a desired output,” said Mark K. Massey, research associate, School of Engineering and Computing Sciences at Durham University.

This emerging field of research is known as “evolution-in-materio,” a term coined by Julian Miller at the University of York.

“Training” CNT-polymer composite materials to perform logic functions (credit: M. K. Massey/J of Applied Physics)

It combines materials science, engineering, and computer science. It uses an approach similar to biological evolution: materials can be “trained” to mimic electronic circuits — without needing to design the material structure in a specific way.

“The material we use in our work is a mixture of [conducting] carbon nanotubes and [insulating] polymer, which creates a complex electrical structure,” explained Massey.

“When voltages (stimuli) are applied at points of the material, its electrical properties change. When the correct signals are applied to the material, it can be trained or ‘evolved’ to perform a useful function.”

The research “could lead to new techniques for making electronics devices [for] analog signal processing or low-power, low-cost devices in the future.”

The research is describe in a paper in the Journal of Applied Physics.


Abstract of Computing with carbon nanotubes: Optimization of threshold logic gates using disordered nanotube/polymer composites

This paper explores the use of single-walled carbon nanotube (SWCNT)/poly(butyl methacrylate) composites as a material for use in unconventional computing. The mechanical and electrical properties of the materials are investigated. The resulting data reveal a correlation between the SWCNT concentration/viscosity/conductivity and the computational capability of the composite. The viscosity increases significantly with the addition of SWCNTs to the polymer,mechanically reinforcing the host material and changing the electrical properties of the composite. The electrical conduction is found to depend strongly on the nanotube concentration; Poole-Frenkel conduction appears to dominate the conductivity at very low concentrations (0.11% by weight). The viscosity and conductivity both show a threshold point around 1%SWCNT concentration; this value is shown to be related to the computational performance of thematerial. A simple optimization of threshold logic gates shows that satisfactory computation is only achieved above a SWCNT concentration of 1%. In addition, there is some evidence that further above this threshold the computational efficiency begins to decrease.