There is a big move towards simulations of brain-like architectures in conventional computers.
View full-size image
The brain, at the centre of the nervous system, is the most complex organ in the body. A vast neural network of around 100 billion neurons makes connections via synapses and is thought to be the basis of memory. Imagine an electronic analogue of a synapse, a memristor, which could be used in making a digital memory for a ‘smart’ computer.
A research team from the University of Leeds, led by Professor Christopher Marrows, aim to make a magnetic memristor which captures the way the brain works, but within a computer. There is the potential to build a circuit which mimics the way neurons are wired together in the brain, presenting a completely different computing architecture.
The material in question is an alloy of two metals, iron and rhodium. Together they form a special ordered structure possessing lots of curious properties, one of which being that the material undergoes a phase transition between two different magnetic states. At low temperatures, the alloy is antiferromagnetic and heating it to the transition temperature induces ferromagnetism in the material.
Properties of iron rhodium can be tuned by doping. Changing a small quantity of rhodium for palladium or iridium alters the transition temperature - iridium increases the transition temperature whereas palladium decreases it.
The Leeds team use neutrons to study these materials. “Neutrons are powerful since they can non-destructively look at what’s going on inside the sample in such a way that you can localise things” stated Christopher, “like the point at which the material starts to become ferromagnetic under certain conditions. Only neutrons can deliver this.”
One of the reflectometry instruments at ISIS, PolRef, was used to look at the alloy at different temperatures. A beam of neutrons was used to bounce off the sample and detect magnetism, as they reflect off the interface where the material changes from an antiferromagnetic into a ferromagnetic state. Therefore the depth of the interface within the sample is known at differing temperatures.
The alloy is a thin film grown on one side of a single crystal magnesium oxide substrate, and is shiny, resembling a mirror, which the neutrons bounce off before they are detected. Through the thin film is a gradient of magnetism dependent on temperature. At room temperature, it should be antiferromagnetic, and when heated to 300C it would be entirely ferromagnetic. At a mid-temperature of around 150C the alloy would be half ferromagnetic and half not, resulting in an interface which can be moved around as a function of temperature.
Christopher Marrows with the spluttering machine that the alloy samples were grown in
View full-size image
The phase transition temperature is around 100C (depending on doping), which is convenient for potential use in magnetic recording business in applications such as heat assisted magnetic recording.
The goal is to build a magnetic version of a memristor. If this can be done the team would then try and show they can move the interface around by either changing the temperature or magnetic field.
“It is interesting because a memristor behaves a bit like the synapses between neurons in the brain and is thought to be how memory occurs” reiterated Christopher. “If a particular connection between two neurons is stimulated, the connection becomes stronger and stronger.”
This is the start of a big move towards simulations of brain-like architectures in conventional computers to understand how they work, and if smart phones really do get a brain, they may be the new PA.
Christopher Marrows, Professor of Condensed Matter Physics at the University of Leeds
Research date: December 2012
For further information, visit University of Leeds Condensed Matter Physics Group
|Other STFC||News||Site Sections||Important Links|