'I t's like throwing spaghetti on a plate." It's not often you hear a scientist describing a technique in such accessible culinary terms, but for Prof John Boland, the process of creating tangles of tiny "nanowires" is that straightforward.
The spaghetti-like wires, which can be as thin as a few nanometres in width and up to tens of microns long, form messy, tunable networks, according to Boland, who reckons they could open up smarter ways of sensing and even computer "learning". And he has just landed a prestigious Advanced European Research Council grant worth almost €2.5 million over five years to work on the tangles.
To make the messy networks, you put nanowires into solution and spray them on to a surface, explains Boland, who is director of the centre for research on adaptive nanostructures and nanodevices at Trinity.
“A few years ago we noticed that when you deposit nanowires, they tend to form these tangled networks. And it is like throwing spaghetti on a plate – you can make a very thin layer of spaghetti or you can make a thick layer.”
When the researchers looked at the properties of these tangles, they saw something interesting. “Rather than having definite properties, these networks tended to have what we call behaviours,” recalls Boland, who is also a professor of chemistry at Trinity.
“Depending on whether the stimulus we applied was electrical or light or chemical, the properties changed.”
In particular, they found that they could tune the ability of the network to conduct electricity.
“We found that the nanowires didn’t all connect at once, rather we could connect some of the nanowires and then leave the others, and that way we could make a material that had some level of conductivity,” he says.
“Then if we wanted to, we could push it forward and connect some more nanowires – so we could choose the conductivity of the network to almost have any value that we want.”
The junctions between the spaghetti nanowires, he explains, are the key.
“We have looked at silver, copper and a range of different semiconductors, and in every case the junctions are what matters. These junctions between the wires are not all the same, they differ very slightly, and because of that there’s a variation in the condition under which it turns on.
“So in contrast to digital electronics where something is on or off, we have materials that can have lots of different levels of being on – from being fully through being fractionally on and then fully off. It is more like an analogue kind of system than a digital one.”
Boland sees near-to-medium-term applications of the nanowire networks as intelligent fuses that can absorb surges of energy, or as smart sensors that could sit on a bridge or building and signal when vibrations are likely to cause damage to the structure.
His team is also looking at applications in solar energy and thermoelectric approaches that interconvert heat and electricity, he adds.
But he has longer-term plans too. “I think the real opportunity is to start to make materials or networks that have the capacity to demonstrate learning.”
Wiring to learn?
In one way, the nanowire networks can be a challenge to work with because they "remember", according to Boland. "They have this incredible memory of what you do to them and if you mistreat them they don't forget.
“When you treat our network, you may turn some connections on and turn some off and you may not be aware of it, but the network will record the fact that you turned them on and off. And we are sure we will be able to make networks that will ultimately show learning characteristics, so we can begin to teach these networks.”
It’s a similar concept to “memristors”, or components that can remember how much current has previously passed through and that influences how much they let through now.
Others are using such approaches to try and build brain-like networks, and Boland reckons the nanowire junctions can do the job too. In that vein he sees the potential for computers that use associative memory, much like the way we do when we are trying to recall something like the name of someone we met at a party.
“The first thing you can see is their face, then you can imagine you can see the first letter of their name,” he says. “The way your brain works is that it associates these things together and then it clicks into place quite rapidly and you home in on the name.”
If computers could work in a similar way, it could mean faster online searches, and better facial recognition by machines, says Boland. "These nanowire materials, engineered in the right way, have the properties that are required to be able to demonstrate associative memory, and that is really exciting for us."
Boland says he is "just about to sign contracts" with the research council for the funding to work on the tangles.
“I am glad to have nailed this down,” he says. “This was a project that was being bootstrapped, but now I can consolidate an existing team, grow it and work with other people in related areas to build up the critical mass.
“The race at the moment is to consolidate the know-how to do this, to make sure we protect the intellectual property and to figure out whether we are going to license these technologies or if there is an option for a potential spin out.”