Algorithm for Robust Correction of Long-Term Drift Components in Gate Leakage Current RTN Data

Investigating oxide defect-induced random telegraph noise signals within stress-induced leakage current is a key approach for understanding the degradation of thin oxides in deeply-scaled devices. However, experimentally measured time-resolved gate leakage current at a constant stress voltage often includes non-constant baseline profiles due to various drift-contributing components. We introduce an algorithm that isolates pure stress-induced leakage current information from gate leakage current data by extracting the underlying baseline profile without prior assumptions. Using a Monte Carlo-generated representative dataset, we demonstrate statistically robust extraction of an arbitrary baseline profile. Subsequently, the algorithm's performance on various baseline profiles is validated, and the significance of baseline correction is highlighted by analyzing random telegraph noise signals in an experimental dataset before and after correction.