Tunable resistance of an oxide thin film can be used to imitate the working of a synapse.
Present day computers need separate circuits for memory and computation. Whereas our brain can both memorize and compute using its cells (neurons). Scientists are trying to develop Neuromorphic Computing Systems that mimic this structure and operation of the brain. However, this typically demands complex input signals and hardware. The neuromorphic computing group led by Prof. Rajendran has now discovered that nano devices using Pr0.7Ca0.3MnO3(PCMO) thin films bypass the need for both.
A neuron carries information via electrical pulses, which travel to a synapse – the junction connecting any two neurons. A key feature of a synapse is that its resistance can change depending on the time difference between the pulses approaching it from the different neurons.
To mimic this behaviour of a synapse, the group fabricated nano devices of PCMO which act like memristors. When a voltage greater than 2V is applied, the resistance of this device increases, taking the device to a high resistance state. On applying such a high voltage again with opposite polarity, the device “remembers” its previous state and switches back to the original low resistance state. Its resistance cannot be changed by a voltage smaller than 2V.
Taking advantage of this behaviour, the team applied input voltage pulses of 1.5V at random times. If the time difference between two pulses is sufficiently low, overlap occurs and PCMO experiences a combination of two signals, sufficiently strong to change its resistance. If time difference is large, individual pulses are not strong enough to modify the resistance and it remains the same. Resistance state is checked by applying a small voltage of 0.5V.
Thus, with only a thin film memristor and a program generating random pulses, IITBNF researchers have been able to demonstrate devices working like synapses.
Work Supported by: Department of Science and Technology and Department of Electronics and Information Technology, India.
Published Paper: N. Panwar, D. Kumar, N. K. Upadhyay, P. Arya, U. Ganguly, B. Rajendran, “Memristive Synaptic Plasticity in Pr0.7Ca0.3MnO3 RRAM by Bio-mimetic Programming” Device Research Conference (DRC), June 2014, 72nd Annual, Page 135-136