Reducing the Impact of Defects in Quantum-Dot Cellular Automata (QCA) Approximate Adders at Nano Scale

Quantum-dot Cellular Automata (QCA) has been studied for some time as a candidate to replace traditional CMOS circuits. QCA circuits are implemented using majority-logic gates and inverters as primitive elements. Different types of defects affecting QCA during synthesis and manufacturing have been identified. Due to the use of the majority logic gate as the main building block of logic designs, the probability and impact of a defect affecting this type of element is significant. The effect can be translated into a change of the boolean expression implemented by the gate. QCA Approximate adders have been proposed for image processing and other applications. This paper analyzes the effect of QCA majority gate defects in the processing of images when using approximate adders. This is done by evaluating the variation in the value of common error distance metrics in the presence of defects. Mitigation of these defects by combining approximate and exact adders and selective introduction of fault-tolerant majority gates in different bits is analyzed. The technique achieves a reduction of around 75% of the average normalized mean error distance (NMED). Specific image-based metrics such as PSNR and SSIM are also evaluated in two different experiments, with an increase of up to 225% and 226% respectively. The increase in area is limited to around 40%.