An oscillator chip inspired by the brain
The figure shows a matrix of 16 oscillators represented by green arrows.
What causes crickets to sing in chorus, fireflies to blink in harmony, and how does the brain store memories? A new dissertation in physics introduces a new magnetic oscillator array through which researchers hope to develop a physical system that imitates the human brain.
If a person is reminded of their first ski trip, a large collection of interacting neurons in the brain start to fire in a synchronous pattern at a rate of several hundred times per second. It is the same pattern that the neurons fired when the person skied for the first time. The pattern remains stable, although the individual neurons may change over time. This is why our memories might fade, but they are never forgotten.
– The ability to memorize is why you only need to see "the fisrt and lsat leteter of the wrods cortecrly placed" to read a sentence. This is because our brain treats words as patterns, not as group of letters, says Mohammad Zahedinejad.
Inspired by how the brain works
In his Ph.D. thesis, Mohammad Zahedinejad was inspired by how neurons in the brain wire and interact. His thesis introduces a magnetic oscillator called spin Hall nano-oscillator that behaves as an artificial neuron.
– The oscillator swings billions of cycles per second, corresponding to GHz frequencies. By exploring different materials, we extended the operating frequency of SHNOs specifically for communication systems. From the range at which today's 4G operators communicate, to tens of GHz frequencies for 5G, all achieved by a single nanoscale device.
Mohammad Zahedinejad and his colleagues hope to be able to create a physical system that can function as the human brain.
– When we place our oscillators close to each other, as in two-dimensional arrays, they can interact and whisper to their neighbors, so that they synchronize and all of them operate at the same frequency. The synchronization frequency, corresponding to the memory, is a unique response of the whole artificial neurons to the stimuli.
Mohammad Zahedinejad’s thesis presents practical approaches to stimulate each artificial neuron by applying an electric field and embedding short-term memory elements, so-called memristors.
Fast cognitive hardware
The hope is that the results of the thesis can lead to ultra-fast brain-like computers that can communicate with each other.
– The impact of the dissertation ranges from understanding the physics of strongly coupled systems to Internet of Things (IoT) where smart machines can learn, decide and communicate over internet, says Mohammad Zahedinejad.
Contact: Mohammad Zahedinejad, 031-786914, 070252972, firstname.lastname@example.org
Supervisor: Professor Johan Åkerman
Thesis title: Spin Hall nano-oscillator arrays: towards GHz neuromorphics
Figure: The figure shows a matrix of 16 oscillators represented by green arrows.
Photo: Mohammad Zahedinejad.