Imagine a crazy watchmaker who, instead of designing springs and gears and planning how to fit these parts together to build a clock, worked using an unconventional method, putting the pieces in a box and shaking it until they fit together to form a mechanism that worked perfectly.
Something similar was obtained by a team from the University of Durham, UK, in collaboration with Brazilian physicist Diogo Volpati, of the Physics Institute, University of São Paulo (USP), in São Carlos. The researchers manufactured a very thin film using a polymer (butyl polymethacrylate) containing a tangle of carbon nanotubes—cylindrical structures formed by carbon sheets only one atom thick rolled in to a tube. This material is of great interest for its electrical and mechanical properties.
When the film was ready, the team applied a sequence of electrical pulses to the material to change the electrical conductivity of the nanotubes and identify how the network could be used to process information. The researchers call this strategy “training” the material. This allowed them to work with the nanotube film to perform a task that only a computer electronic circuit can carry out: process electrical signals in order to perform logical operations.
“This is a new approach to manufacturing electronic nanodevices, on the boundary between computer science, materials science and electrical engineering,” says Volpati. “Instead of assembling electronic circuit boards to process information, we ‘train’ an initially unorganized material to perform the desired task.”
This way of producing new materials that behave like electronic circuits was inspired by how living organisms evolve. In 2004 physicist Julian Miller, of the University of York, UK, coined the name evolution in materio for this strategy. It allows the creation of electronic circuits without the need for full control over the assembly of the circuit’s structure. Physicists, engineers and computer scientists believe that evolution in materio is one way to overcome a problem that haunts microelectronics: the limit on miniaturization of computer chips. In recent decades the ability to process information has increased continuously because more electronic circuits have been carved into increasingly smaller spaces. But it is reaching the point where manipulating matter in order to further shrink circuits is no longer physically possible.
Miller and other researchers are pursuing this goal as part of the Nanoscale Engineering Design for New Computing Using Evolution Project (Nascence), which involves five European universities and, since 2012, has been promoting collaboration between computer scientists, physicists and engineers in search of new materials and different ways to train them.
A specialist in the manufacture and characterization of new nanomaterials, Volpati was asked by Michael Petty, leader of the Durham group participating in the Nascence project, to collaborate on the experiments with the material made of carbon nanotubes mixed with the polymer. Under certain conditions, the nanotubes can play the role of microscopic electrical wires. Thus, the tangle within the polymer acts like an electronic circuit, but it is a mess before the experiment begins.
In tests, a small piece of the material is connected to normal computers through a series of electrodes. The function of some of them is to trigger electrical pulses representing the input data—a sequence of numbers, for example—for a mathematical calculation. These pulses traverse the network of carbon nanotubes and are captured at the other end. The output pulses correspond to the solution to the mathematical problem.
However, at the beginning of the experiments the network scrambles the input electrical signals because the nanotubes are disordered. The result is that the output data gives the wrong answer to the problem. The ability to solve the problem improves as additional electrodes trigger electrical signals produced by a computer program whose purpose is to identify the configuration of electrical pulses able to modify the spatial orientation of the nanotubes. Performed via trial and error, the work of this program, which uses what is known as an evolutionary algorithm, takes only seconds and allows researchers to discover the best way to align the nanotubes and process certain pieces of information in a set of electrical circuits with an unknown structure.
Using this strategy, Mark Massey, a postdoctoral researcher at the University of Durham, Volpati and their colleagues tested different mixtures of polymers with carbon nanotubes. In an article published in April 2015 in the Journal of Applied Physics, they showed that the material only performs a given mathematical operation if the concentration of carbon nanotubes dispersed among the polymer molecules varies between 0.11% and 1%. “The material has the physical properties necessary to accomplish the task only in these concentrations,” explains Volpati.
The computation performed by the material in this experiment was very simple. The tangle of nanotubes carried out only three types of sums: 0 + 0; 1 + 1; and 1+ 0. In 2014, however, Petty’s team had already used the same material to resolve a slightly more complex problem, known as the Traveling Salesman Problem: determine the shortest route a traveling salesman should follow to visit a series of neighboring cities. The nanotube film was used to solve the problem for up to 12 cities arranged in a circle on a map. “These are proof of principle,” explains Volpati. “The challenge now is to develop a piece of material able to replace the electronic circuit board controlling a robot, for example.”
One of the difficulties that researchers face is that carbon nanotube films are rigid, with a serious limitation: learning is fleeting. They only act as electronic circuits while the pulses of the evolutionary search algorithm are being applied. Once these pulses are turned off, the material loses the electrical properties that allow it to function as a circuit.
In another article published in 2015 in the Journal of Applied Physics, Volpati and his colleagues described a possible evolution of this material. The researchers were able to replace the polymer’s rigid matrix with one made of liquid crystal. Unlike the polymer molecules, which do not move, the liquid crystal molecules move under the influence of the electric pulses emitted by the evolutionary algorithm. “The liquid crystals change the spatial orientation of the nanotubes, permanently altering the properties of the material,” explains Volpati. “We also show that, after orienting the nanotubes in the direction we want, they stay put and the material does not lose its memory.” Currently, the researchers are trying to use the nanotubes immersed in the liquid crystal to perform mathematical operations.
1. Spectroscopic evaluation of the bulk and interfacial molecular orientations of organic thin films deposited onto different surfaces (No. 2012/09905-3); Grant mechanism: Post-doctorate research grant Principal investigator: Osvaldo Novais de Oliveira Junior (IFSC-USP); Grant recipient: Diogo Volpati; Investment: R$168,972.87 (FAPESP)
2. Molecular control in nanostructured films of carbon nanotubes (No. 2013/08864-4); Grant mechanism: Post-doctorate research grant abroad; Principal investigator: Osvaldo Novais de Oliveira Junior (IFSC-USP); Grant recipient: Diogo Volpati; Investment: R$202,700.20 (FAPESP).
MASSEY, M. K.et al. Computing with carbon nanotubes: Optimization of threshold logic gates using disordered nanotube/polymer composites. Journal of Applied Physics. V. 117, No. 13. April 6, 2015.
VOLPATI, D. et al. Exploring the alignment of carbon nanotubes dispersed in a liquid crystal matrix using coplanar electrodes. Journal of Applied Physics. V. 117, No. 12. Mar. 24, 2015