Application of robust fitting analysis techniques to low-resolution spectral data

Abstract

A recently developed spectral analysis methodology, RobWin, is applied to low-resolution spectra from room-temperature semiconductor instrumentation to get a first look at its ability to identify the presence of nuclides in these spectra. The method, which builds upon non-linear robust fitting techniques from RobFit spectral analysis software, emphasizes adding structure gradually to the entire spectrum background shape function after accounting for re-optimized combinations of full-spectrum photopeak functions for a user-defined set of nuclides at each iteration. This feature makes RobWin well suited for the identification of weak-strength nuclides in high-background environments. The analysis approach emphasizes modeling the continuum for the entire spectrum as a single, continuous curve composed of cubic splines with optimized knot locations and coefficients. Peak shapes can be derived from actual data, a feature particularly useful with room temperature semiconductor materials. In its peak-search mode, RobWin can choose the best match from up to 10 different peak shapes for each unidentified peak. The data were collected with CZT-based detectors and were provided by the IAEA so that an initial look at the use of the RobWin methodology for analysis of room-temperature semiconductor spectra could be performed. The analysis methodology is briefly described and initial results are presented.