Background subtraction using the SNIP method enhanced by the Levenberg–Marquardt algorithm
Anh Nhi Le, Thanh Tai Chau, Ngoc Ba Vu, Thien Thanh Tran, Le Hao Cong, Van Tao Chau
Nuclear Instruments and Methods in Physics Research Section a Accelerators Spectrometers Detectors and Associated Equipment, 171544 (2026)
Abstract
In this study, a practical refinement of the SNIP method is proposed for background subtraction in HPGe γ-ray spectra. A single filter width determined from the detector specifications is used in Gaussian convolution to obtain preliminary peak parameters, which are then refined using the Levenberg–Marquardt (LM) algorithm to provide more reliable peak widths for SNIP background subtraction. The resulting method, referred to as LMSNIP, is evaluated using the 154Eu spectra measured at 25 cm and 5 cm. The reconstructed backgrounds are directly compared with those obtained using the adaptive SNIP approach, while the resulting net peak areas are compared with Genie2K. For the 25 cm spectrum, LMSNIP and adaptive SNIP provide generally similar background estimates. Under the more demanding 5 cm condition, however, LMSNIP yields a smoother and more stable baseline and generally better agreement with Genie2K for most investigated peaks. Additional validation using the 133Ba spectrum measured at 5 cm further shows that LMSNIP remains effective for a reference spectrum with different low-energy characteristics. These results support the applicability of LMSNIP to background reconstruction and net peak-area determination under the investigated close-geometry conditions.
More detail >>.
