Scientists are trying to emulate the human brain since it is the ultimate computing machine. In this effort, the latest research has resulted in the development of a device which can also “forget” memories much like our brains.
It is known as a second-order memristor. It mimics the synapse of a human brain in such a manner where it stores information but then loses it slowly when it is not accessed for a long time period. The device currently does not have a practical use but this could be a stepping stone to a unique kind of neurocomputer which can perform the same functions that a human brain does. The work appears in ACS Applied Materials and Interfaces.
In an analogue neurocomputer, neurons and synapses can be replicated by the on-chip electronic components. This could help in amplifying computational speeds as well as decreasing the energy requirements of the computer.
Presently the analogue neurocomputers are not feasible as researchers need to figure out how synaptic plasticity can be also implemented in electronics. This is the technique in which the active brain synapses become strong while the inactive ones get weak resulting in fading away of memories.
Previously, memristors were produced by nanosized conductive bridges which decayed with the passing of time similar to how we forget some incidents.
Anastasia Chouprik, a physicist from the Moscow Institute of Physics and Technology(MIPT), Russia said that in the first order memristor, the problem is that the device changes its behaviour with the passage of time resulting in its breakdown. The synaptic plasticity has been implemented in a robust manner this time which sustained the change in the state of the system for 100 billion times.
A ferroelectric material, hafnium oxide was used along with electric polarisation which changes in response to an electric field. It is already used by Intel for manufacturing microchips. So it would be easier to introduce the memristors.
Researchers faced challenges in finding the proper thickness for the ferroelectric material. They found four nanometres to be the ideal thickness as a nanometre more or less would make it unsuitable for application.
The forgetfulness is implemented through an imperfection as a result of which microprocessors based on hafnium are difficult to develop. The imperfection is the defect present at the interface between hafnium oxide and silicon which results in the decrease in the memristor conductivity.
There is a long way to go as these memory cells have to be made more reliable and suitable enough to be integrated into flexible electronics. Another physicist, Vitalii Mikheev said that they would be studying the relation between several mechanisms through changing the memristor. There might be mechanisms other than ferroelectric effect which have to be studied.
Journal Reference: ACS Applied Materials and Interfaces