Efficient disposable chips from organic materials

Calculating and processing information the way our brain does – that would mean a huge revolution for everything we now use classical computers for. This so-called neuromorphic computing is much more efficient and therefore much more energy efficient. And that makes completely new applications possible, as can be read in part one of this diptych.

One with your body

But there are many more possibilities if you try to get closer to the biological example. Because if you can make ‘computers’ that really work like our brains, you can also communicate much more directly with biological materials, with living cells and even with our nervous system. This would enable prostheses that not only understand signals from our nerves, but also provide feedback the other way around. This allows our brain to train and control the prosthesis better, so that the attachment becomes much more one with the body.

You can also think of combinations of sensors that you place in or on the body and that continuously measure things such as heart rate, blood sugar level or oxygen level, and can immediately deduce from this whether, for example, a medicine needs to be delivered. And because such a ‘smart patch’ does all the calculations itself, there is no need for an exchange with a central computer elsewhere. It works completely independently. That makes a huge difference: no wiring, no batteries and no need to be connected to a network.

But if you’re serious about getting such a brain imitation, you’ve got to go “biologically”. As Yuri van de Burgt, assistant professor at Eindhoven University of Technology. He is also working on neuromorphic chips, but from a very different starting point than the chips by Wilbert van der Wiel and colleagues, which are very innovative, but also very conventional in the materials they use. Van de Burgt opts for soft, organic materials and hopes eventually to make cheap, easy to train and easily printable chips – almost towards disposable chips, and then you don’t have to work with silicon and gold.

Test setup of a neuromorphic chip in the lab of Yoeri van de Burgt. PhD candidate Eveline van Doremaele measures how incoming signals are converted into outgoing signals.

Esther Thole for NEMO Kennislink

Carbon as base

The living world does not run on gold or silicon, but on organic materials consisting of carbon, hydrogen, oxygen and nitrogen. That is why Van de Burgt uses these building blocks to build his brain-inspired chips. This concerns so-called conductive organic polymers; these are long molecules that can pass an electric current and that adjust their resistance as soon as negatively (electrons) or positively (ions) charged particles enter or leave the material. “These polymers are neutral. If a positively charged ion enters the material, it must be compensated. In response to this, the polymer pushes out a negative charge, an electron, which creates an adapted resistance,” explains Van de Burgt. “You can see the adjusted resistance as a ‘memory’, the material remembers that change.”

Much communication in our body and brain takes place via ions. Materials such as gold and silicon can do nothing with such a signal, but organic polymers can. This makes them very suitable for not only imitating the functioning of brain cells, but also for communicating directly with living cells. Van de Burgt and his group demonstrated this by linking a small electrical circuit, with a layer of a conductive polymer on top, to living brain cells. These cells secrete the chemical dopamine, which is an important signal in our brains. “The released dopamine reacts with oxygen and an electron from the polymer layer is also used in that reaction. The polymer then adjusts its resistance to remain neutral, resulting in an altered conductivity. You can measure that as a changing current that flows through the circuit.” For example, the team showed that you can control an artificial electronic system with a natural signal from living cells.

Stronger synapse

Not only are the materials and signals used differ from the ‘hard’ neuromorphic chips, the processing of the signals into information is also fundamentally different. The chips that Van de Burgt makes do not require an external processor for processing: everything is done immediately in the hardware, or the organic material. The material is in fact already ‘programmed’ to respond to signals in a certain way. “What we do would you in-memory-computing can name. We create something that resembles a neural network: a chain of interconnected nodes, in which each connection is given a certain ‘weight’ that indicates how important this connection is for the end result. That is very similar to how things work in the brain. Neurons are connected via a so-called synapse (a connection) and the more often two neurons communicate with each other, the stronger that synapse becomes. This way you can also see the connections in our network: the more important a connection is for the end result, the more weight it gets.”

Detail of the measurement setup of PhD student Eveline van Doremaele. The needles pierce a layer of the organic, conductive polymers. An incoming electrical signal travels through the layer to another part of the chip, where the outgoing signal is measured via another needle.

Roel van der Heijden for NEMO Kennislink

The big advantage, so far in theory, is that such an in-memory system can be trained much more efficiently. “It’s a network of resistors that you adjust, all at the same time. That is very efficient.” Thanks to the organic polymers, programming connections in such a network is very predictable, says Van de Burgt. “An ion in it is an electron out. That does not change.” A drawback of the organic polymers is that they not only react to the desired signal, but also react with, for example, oxygen from the air. “Such parasitic reactions are inevitable; organic materials react very easily. As a result, the memory of our networks does not last indefinitely: over time things are ‘forgotten’.”

Nevertheless, Van de Burgt sees important opportunities for these organic chips. “Due to the high predictability, these chips can work very autonomously. You know exactly what they’re doing and you don’t need an external computer to check whether what you wanted to program actually worked. And you can train such a network over and over again, which is not possible with a silicon chip.” Other advantages, he says, are that these systems consume very little energy, that they are soft materials and that they are biocompatible: they can directly ‘talk’ to biological materials. The fact that they are easy to make and print and are also very cheap compared to the materials in the classic chips is of course also a plus.

Train again

The big question is, of course, what you can use these organic chips for in the long run. They do not appear to be a serious competitor for the chips of giants such as IBM and Intel. That is not what Van de Burgt has in mind either. “We see lab-on-a-chip applications (miniature versions of laboratory tests, ed.) as a good niche, for example for diagnostics. Such a chip can classify signals in order to obtain information about a possible disease – such as recognizing cancer cells in a blood sample. The advantage of our chips would be that you can always process new insights in the chip, because you can train it again on the basis of new data.”

Schematic representation (not to scale) of the communication between a nerve cell that receives biological signals and releases dopamine in response. The dopamine then reacts and is converted into an electrical signal by the neuromorphic material.

According to Van de Burgt, another important application lies in direct communication with our body. Think of sensors that you stick to the skin and that continuously register and process signals from your body and on the basis of this they deliver medicines. Or implants, such as pacemakers or electrodes in the brain, which are then no longer made of rigid materials that are foreign to the body, but of materials that fit in much better with our body. “Because our chips consume so little energy and all information processing takes place directly in the chip, you do not need a connection to external devices or power supply.”

Finally, Van de Burgt sees opportunities to allow prostheses to communicate better and more directly with the brain by coupling the nerve endings to an ‘interface’ of conductive polymers. “If you make this intermediate layer adaptive, so that the material adapts under the influence of signals from the brain, the patient can train the prosthesis much more directly. This is much faster than if the prosthesis first has to send the signals to an external computer, which processes everything and returns a corrected signal. I think we can offer many benefits here with our materials.”

Views

The possibilities are there, but the practical applications are not yet in sight, admits Van de Burgt. “The big challenge for us is how we can train such a network properly and efficiently by adjusting the weight of all connections at once. If we understand that principle, we can really take a step.”

Sources:


Source: Kennislink by www.nemokennislink.nl.

*The article has been translated based on the content of Kennislink by www.nemokennislink.nl. If there is any problem regarding the content, copyright, please leave a report below the article. We will try to process as quickly as possible to protect the rights of the author. Thank you very much!

*We just want readers to access information more quickly and easily with other multilingual content, instead of information only available in a certain language.

*We always respect the copyright of the content of the author and always include the original link of the source article.If the author disagrees, just leave the report below the article, the article will be edited or deleted at the request of the author. Thanks very much! Best regards!