As emerging non-volatile memory (NVM) devices, Ferroelectric Field-Effect Transistors (FeFETs) present distinctive opportunities for the design of ultra-dense and low-leakage memory systems. For matured FeFET manufacturing, it is extremely important to have an understanding of manufacturing defects and accurately model them to develop effective test solutions. This paper introduces a comprehensive framework for defect and fault modeling, which enables the development of test solutions. First, a classification of FeFET manufacturing defects is provided; both conventional defects (such as contacts and interconnect defects) as well as unique FeFET defects are discussed. The latter FeFET specific defect leads to unique faults that cannot be adequately described using traditional modeling approaches. Then, the Device-Aware Test (DAT) method is used to effectively and appropriately model, analyze and develop test solutions for such unique defects; the approach will be illustrated for Stuck-at-Polarization (SAP) defects.

The mobile element variant is a very important structural variant, accounting for a quarter of structural variants, and it is closely related to many issues such as genetic diseases and species diversity. However, few detection algorithms of mobile element variants have been developed on third-generation sequencing data. We propose an algorithm ricME that combines sequence realignment and identity calculation for detecting mobile element variants. The ricME first performs an initial detection to obtain the positions of insertions and deletions, and extracts the variant sequences; then applies sequence realignment and identity calculation to obtain the transposon classes related to the variant sequences; finally, adopts a multi-level judgment rule to achieve accurate detection of mobile element variants based on the transposon classes and identities. Compared with a representative long-read based mobile element variant detection algorithm rMETL, the ricME improves the F1-score by 11.5 and 21.7% on simulated datasets and real datasets, respectively.