A Novel Defect Tolerance Scheme for Nanocrossbar Architectures with Enhanced Efficiency

Abstract

The semiconductor industry is now facing challenges to keep pace with Moore’s law and this leads to the requirement of new materials and newer technological devices. Molecular switch-based nanodevices are one of the promising areas because of their ultimate size and miniaturisation potential. These nanodevices are built through a self-assembled bottom-up manufacturing method in which the possibility of external intervention is negligible. This leads to a considerable yield loss due to defective device production and the traditional test-and-throw faulty device approach will not hold well. Design of fault-tolerant devices are the only possible solution. A widely studied nanodevice is nanocrossbar architectures and their fault tolerance can be designed by exploiting the programmable logic array’s fault tolerance schemes. A defect-unaware fault tolerance scheme is developed in this work based on the bipartite graph analogy of crossbar architectures. The newly-designed algorithm can eliminate more than one node in each iteration and, hence, a defect-free subcrossbar can be obtained much faster compared to the existing algorithms. A comparison with the existing defect-unaware fault-tolerant methods with this newly-developed algorithm shows a better yield in most of the cases.