Neuromorphic engineering is an interdisciplinary field of applied sciences that combines biology, physics, mathematics, and neuroscience to develop hardware models of neural and sensory systems. Our primary focus in CenBRAIN in this field is the design of neuromorphic chips that utilize analog and mixed-signal techniques as well as memristors.
Figure 1. A neuromorphic chip concept.
Architecture of neuromorphic computing units is bioinspired by the biological brain. It aims at implementing artificial neural networks in a chip. Functional building block of such computing units are neurons, axons, synapses and dendrites. Among them, synapses connect neurons and must be able to remember previous state, updates to a new state, and holds the weight of the connection. Computation and data storage take place at the same place. Neuromorphic computing includes the production and use of neural networks and design of low power integrated circuits to achieve highly parallel non von Neumann architectures.
Implementation of transistor-based synapses doesn’t match the expected neuromorphic chips due to their bulky geometry, power-hungry operation and slow operation. Memristors are promising candidates to implement these artificial synapses due to their CMOS compatibility, high-density devices, fast-switching speed, and low-power consumption. Besides, metal oxides and metal-organic complexity, the molecular organics (molecules and polymers) are promising contenders (readily available low cost starting materials, ease of purification, roll to roll processibility, and mechanical flexibility) with resistive switching characteristics. Modern synthetic techniques offer a rich opportunity for manipulating the architecture and electronic structure of organic compound molecules, enabling organics-based resistive memory to provide desired electronic performance. Using CenBRAIN’s platform, we develop novel materials with structurally tuned molecular entities that sync with memristor architectures.