Abstract

The basis of the powerful operation of the brain is the variability of neuron operation, i.e., it can be digital or analog, and the logic operation of a neuron-based system can be parallel and series. The challenge is to set up an artificial intelligent architecture that mimics neuro-biological architectures present in the nervous system. Our proposed new active device (phonon transistor = phonsistor) and thermal electric logic circuit (TELC) seem to be appropriate devices for neuron modeling. Key elements of the phonsistor and TELC are memristors realized by VO₂ phase change output resistors integrated with dissipating elements as inputs. These components are coupled to each other by thermal and/or electrical effects. On short distances, the information can be carried by heat diffusion and on longer distances by electrical signals. This is a similarity with human neurons where the information is carried by diffusing neurotransmitter molecules on short distances and electrically by the axons on longer distances. For example, very new ideas are presented of neuromorphic circuits for mimicking the biological neuron synapse operation and the action potential generation. Further similarities with biological neural systems are the auto-oscillation phenomenon with chaotic properties, the ability of integrating subthreshold excitations within the thermal time constant, and the memory effect of the memristive components. The TELC should be compatible with CMOS technology, as the realization of both systems utilizes conventional thin-film technology steps at similar temperature ranges. The physical appearance and construction of the TELC gate are also similar to the neuron.