Researchers have created a substance that can mimic the way the brain stores data. The material mimics learning that takes place during deep sleep by simulating the connections of neurons. The neuromorphic computer technique was used by the research team from the Universitat Autnoma de Barcelona (UAB) to create the magnetic material.
Neuromorphic computing is a computing concept that uses artificial neurons to mimic brain behavior and the synaptic functions or communication signals of neurons. A mimic of brain function is neuronal plasticity. Which is the “ability to store or forget information dependent on the duration & repetition of electrical impulses that stimulate neurons,” according to the study. This form of plasticity is related to memory and learning in the brain.
Materials that emulate learning
The research team discovered certain materials that mimic neuronal synapses. Materials include memresistive (electronic memory) materials, ferroelectrics, phase change memory materials, topological insulators, & magneto-ionic materials. The team found that magnetionic materials were the newest materials and are formed by changes in magnetic properties caused by the movement of ions or atoms within the material.
Motion is created by applying an electric field to the ions. For magnetionic materials, researchers know how magnetism is controlled when an electric field is applied, but it is difficult to control progression of magnetic properties when the voltage is stopped. This makes it difficult to mimic how the brain works, such as the learning process that occurs even when the brain is in deep sleep and not receiving external stimulation.
The study was led by researchers from the Physics Department of UAB Jordi Sort and Enric Menéndez in collaboration with the ALBA Synchrotron, the Catalan Institute of Nanoscience and Nanotechnology (ICN2) and the ICMAB. The team proposed a novel way to control the evolution of magnetization in the stimulated & post-stimulated states of brain function.
A new material with a thin layer of cobalt mononitride (CoN) has been developed. When an electric field was applied, The accumulation of N (nitrogen) ions in the line between the layer and a liquid electrolyte could be controlled. “The new material works with the movement of ions controlled by electrical voltage, analogous to that of our brain, and with speeds similar to those of neurons, on the order of milliseconds,” explains Jordi Sort, research professor at ICREA and Enric Menéndez, Serra Húnter Professor at the UAB Department of Physics.
“We have developed an artificial synapse that could be the basis of a new computing paradigm in the future, an alternative to that used by current computers,” Sort and Menéndez continued. The team found they could emulate processes like memory, information processing, and information retrieval by applying voltage pulses. In addition, for the first time they could mimic the controlled updating of information without the applying voltage. The controller used in the study was created by modifying the thickness of the cobalt mononitride (CoN) layers, which determines the rate of ion movement and the frequency of the pulses from volts.
The material’s configuration permits the magnetoionic characteristics to be regulated both when the voltage is supplied and when it is removed. Once the voltage stimulus has recedes, the magnetization can be increased or lowered depending on the thickness of the material and the method of voltage was applied.
The results from the study
The new material breakthrough opens up a whole new world of possibilities for neuromorphic computing functions, improving perception, learning, and memory through the use of neural networks. The possibility of replicating neuronal learning that takes place during sleep after bran stimulation is illustrated by the example given. The functionality cannot currently be duplicated by any other neuromorphic material.
“When the thickness of the cobalt mononitride layer is less than 50 nanometers and a voltage is applied with a frequency of more than 100 cycles per second, we managed to emulate an additional logic function,” say Sort and Menéndez.
Researchers mentioned the importance of studying and mimicking brain function. “Once the voltage is applied, the device can be programmed to learn or forget without the need for an additional input of energy, by mimicking the synaptic functions that take place in the brain during deep sleep, when information processing can continue with out applying any external signal” explained Sort and Menéndez.
The study was published in the journal Materials Horizons.