A REMARKABLE electronic circuit made entirely from the cells of an LCD display has been created using software that mimics evolution鈥檚 drive for fitness. By reorienting the liquid crystals in individual cells, the software can transform a display into a circuit that can distinguish between high and low-frequency signals, without using any conventional electronic components like transistors, capacitors or resistors.
The circuit was developed by electronics engineers Simon Harding and Julian Miller at the University of York in the UK. It emerged from an experiment aimed at discovering whether a genetic algorithm (GA) 鈥� software that mimics natural selection to find the 鈥渇ittest鈥� solution to a problem 鈥� can be used to exploit the electrical properties of certain materials to generate electronic circuits.
Liquid crystals looked promising because they not only conduct electricity but can also, in certain orientations, act as capacitors. This means an alternating current can pass between two neighbouring crystals even if they are not actually in contact.
Advertisement
Liquid crystal displays are also electronically reconfigurable. Applying a potential difference to an LCD cell creates an electric field that changes the orientation of the liquid crystals within it. The researchers hoped that by twisting individual crystals a few degrees they could change their electrical properties in ways that could be exploited to produce a useful circuit.
In their experiment, Miller and Harding used GA software running on a PC to turn part of a 180-by-120 pixel monochrome LCD into a circuit able to discriminate between signals representing tones of different frequencies. When the circuit is fed two signals, one after the other, it produces an output signal representing a digital 1 if the first tone had a higher frequency than the second, and a digital 0 if the second had the higher frequency.
To produce their circuit, they applied randomly chosen voltages to each of up to 64 pixels of the LCD. For each pattern of applied voltages, they tested the circuit by applying two tones of different frequencies to one pixel. Some 40 voltage patterns were tried. Those that most resembled a tone discriminator were assigned high fitness scores. All the circuits were then 鈥渕utated鈥� by changing the voltages applied to the cells, those with lower fitness being given more mutations than those with higher. After the mutation process had been repeated 100 times, an arrangement capable of reliably acting as a tone discriminator had emerged. However, they have no idea why it works, just that it does.
Although the liquid crystal processor works under controlled lab conditions, Harding and Miller say it has not yet been shown to be robust enough to work at different temperatures and with less stable voltages. Once they have done that, they plan to try to produce a liquid-crystal-based obstacle avoidance system for a robot.
Genetic algorithms are already a standard tool in engineering, where they are used for tasks like optimising the mix of metals when formulating alloys. But when used in electronics, their power is restricted by the limited range of characteristics of devices such as transistors, which act as switches and are either on or off. This limits the possibilities that GAs come up with, the researchers argue. Liquid crystals, on the other hand, are rich in physical properties that can modify electric currents.
Harding and Mitchell say that their work shows it might one day be possible to make novel, reconfigurable circuits by altering the physical and electrical properties of a single material rather than by using discrete components. In addition to liquid crystals, they think some colloids, electroactive polymer molecules, and even bacteria could be used to make such circuits.
