Determination of a mixture of gamma-emitting radionuclides using solid scintillation detectors and multivariate calibration Original Research Article

Solid scintillation detectors are more efficient than germanium detectors, and thus can be used to quantify gamma-emitting radionuclides with shorter counting times. However, due to the low peak resolution of solid scintillation detectors, quantification based on the straightforward calculation of photopeak areas becomes extremely inaccurate for mixtures of radionuclides with photopeaks of similar energies. In this paper, two multivariate calibration methods, partial least squares regression (PLS) and multivariate linear regression (MLR), have been investigated for their capacity to determine the activity of gamma-emitting radionuclides with close photopeaks (85Sr, 134Cs and 110mAg) in absence and in presence of potential interferences by using solid scintillation detectors. The influence of different calibration sets on the prediction of activities has also been tested. The prediction errors obtained when applying either PLS or MLR to the samples without interferences are generally lower than 4% and they are comparable with the errors obtained with germanium detectors. However, when MLR and PLSR are applied to samples with interferences, partial least squares is clearly revealed as the method of choice, giving accurate quantitative results regardless of the identity and amount of other isotopes present in the samples analysed.