Defect-tolerant digital filtering with unreliable molecular electronics

Abstract

Molecular electronics such as silicon nanowires (NW) and carbon nanotubes (CNT) are considered to be the future computational substrates due to their ultra-high density and superior energy efficiency. However, excessive defects from bottom-up self-assembly fabrication pose a major technological barrier to achieving reliable computing at the molecular scale. Existing solutions targeting absolute correctness introduce high cost and complexity in post-fabrication testing and defect diagnosis. In this paper, we propose a new approach exploiting algorithm level enhancements for defect-insensitive signal processing. By deliberately allowing molecular-scale integrated systems to bear defects, the proposed design framework achieves reliable signal processing while significantly reduces the cost of defect tolerance.