This computer chip can make – and erase – memories like a human brain

Using light-responsive, semi-conductive material, RMIT engineers have devised a chip that stores information like the human brain makes memories.

The human brain is an incredibly powerful and complex organ, with computing power far in excess of the most powerful machines in existence today.

Estimates suggest it has more than 100 billion neurons and 160 trillion synapses. These are the connections between which electrical signals fire, creating the sensations that make up thought.

Now, engineers from the Royal Melbourne Institute of Technology (RMIT) have replicated those processes on an electrical chip, creating circuitry that can store and erase information much the same way the human brain is able to make — and forget — memories.

“These neurons, they are connected in networks and they communicate to each other,” RMIT Research Fellow Taimur Ahmed told create.

“If you look at two neurons, when they bond to each other, they start to make a memory. And when, over time, or due to different reasons, such as neural disorders, when this bond becomes weakened or there’s a disconnection between the bond, the human brain starts to lose this memory.”

Mimicking this process in an electronic chip would not only allow engineers to devise radically smaller computer storage systems, it would also give neurological researchers new ways to study disorders of memory and the mind, such as dementia or Alzheimer’s disease.

“Both of those are interlinked,” RMIT Senior Lecturer and team leader Dr Sumeet Walia told create.

“In the studies that we did, we were actually more focused on the neural behaviour. But in the past few years, back when we started on this, it actually started off as more of a miniaturised and high-density data storage technology.”

Walia and Ahmed’s technology makes ingenious use of an atomically thin light-sensitive material called black phosphorus.

Exposing black phosphorus to light allows it to record and destroy information. (Image: RMIT University)

“When we were working with this material, we came across a very unique electrical response to different colours of light,” Walia explained.

“We then thought, okay, you know, this is very unique, and we could use it to mimic neural connections as well as disconnections.”

This ability to mimic disconnections is vital to the technology; it is far more difficult to imitate what neurologists call a depression than it is to create a potentiation — a connection.

“If we … expose this material to ultraviolet light, then it will give us a positive photocurrent,” Ahmed said.

“And if we shine the same material with a longer wavelength — for example, a blue wavelength, or green wavelength — then it will give us a negative electrical pulse.”

Black phosphorus can work like this specifically because of defects in its surface — something that would usually present a problem for optoelectronics.

“What defects can do is act like trap sites for your charge carriers,” Walia said.

“What happens in general defect-free material is, when you shine light, any current generated because of the light will always go up — so you will increase the current being generated because you’re supplying external energy to the material. But when you have certain types of defects present, and couple a certain wavelength of light, these defects actually trap those charge carriers. So, for particular colours of light, your current will actually go down.”

Ahmed said that one advantage black phosphorus has over other super-thin materials like graphene is that it is not as conductive.

“You can change the band gap of the material and you can use it as a semiconductor,” he explained.

Dr Sumeet Walia (left) and Dr Taimur Ahemd. (Image: RMIT University)

Although the technology is still in an early stage, Walia said it could eventually shrink digital storage devices to a fraction of their current size.

“If you can use much smaller materials, but pack in more data, you will get a higher density storage,” he said.

“The holy grail in terms of getting there is getting petabytes on a pinhead.”

The technology could also help neuroscientists better understand how the brain works.

“There are a lot of unknowns about the brain,” Ahmed said.

“For example, to study the functions of a brain or the functions of a neuron, what usually neuroscientists do is that they will do open surgery of a brain, and they will take out the cells of the brain, and then they will put those cells or neurons in a test tube or a Petri dish.”

Even less invasive approaches, such as those used to help people experiencing memory loss, present problems to researchers.

“They expose the brain to electromagnetic radiation, for example, and that electromagnetic radiation will affect the whole brain,” Ahmed explained.

Using light to target specific neurons, however, allows researchers to modify or control those neurons’ function.

“You can have something like a brain on a chip, which you can use to understand these neural disorders better,” Walia said.

“That will give us, or neurobiologists or neuroscientists, so much more flexibility in trying to do bolder experiments and understand how the brain behaves at a fundamental level.”

“These defects actually trap those charge carriers. So, for particular colours of light, your current will actually go down.”
Dr Sumeet Walia

Ahmed emphasised that the research is still at a very preliminary stage. The next step for the team is to scale up the technology in a cost-effective way.

“And also looking at the energy requirements in terms of how much energy this particular chip will be consuming,” Walia said.

“From quick rough calculations, the energies that it’s consuming are fairly low, but to put a number to it is also important.”

The bionic brain

If an electronic chip could store memories the way a human brain does, could the same technology one day be used to create a bionic brain — a system of circuits capable of thinking for itself?

The chip can imitate the workings of the brain. (Image: RMIT University)

Walia said his project could lead to artificial intelligence (AI), but that is a long way away at the moment.

“I think the behaviour of mimicking the neural synapse is more critical as a first step to getting there, because you need smartness in the material,” he said.

“If you can mimic the learning capabilities of the brain, that’s where the seed of AI comes from. And then, of course, because there’s a lot of data to process, the storage follows up as the next step. But if you can mimic the behaviour of neurons, that is the first step towards what’s touted as an eventual bionic brain.”

The brain, he cautions, is a complex organ.

“It’s not as simple as just simply mimicking neurons. But this is the first step in that direction.”

Are you an innovative engineer? Know someone who is? Register your interest for the Most Innovative Engineers listing. create will call for nominations in January 2020. 

Exit mobile version