Chemometric approach to discrimination and determination of binary mixtures of silver ions and nanoparticles in consumer products by graphite furnace atomic absorption spectrometry.

Extensive use of nanomaterials in consumer products inevitably leads to their release to the environment. For monitoring of the fate of nanoparticles in various matrices, low-cost, simple and rapid size-based methods are required. Graphite furnace atomic absorption spectrometry (GFAAS) can be used for this purpose, because due to the difference between a thermal energy required to atomize ionic silver (Ag+) and silver nanoparticles (AgNPs) of different size, the maxima of signals measured by are slightly shifted in time. However, signals are seriously overlapped. Analytical signals of various concentrations of Ag+ and differently sized AgNPs in their mixtures were registered under optimized temperature conditions using GFAAS. The research aim was to develop a procedure for quantitative and qualitative analyses of Ag species in the binary mixtures without the need for the deconvolution of their signals into single components. For evaluation of the qualitative composition of the silver forms present in the studied samples, soft independent modeling of class analogy (SIMCA) was applied. The best results were obtained based on the first derivative of absorption signal. The correct identification rate of the Ag forms was only in the range 25-60% depending on the composition of the tested samples, that indicates limitations of this approach. For estimating the concentrations of the Ag forms, partial least squares regression (PLSR) was implemented. The errors levels (expressed as the root mean squared error of prediction of concentration) for mixtures were acceptable and comparable between the studied databases. The developed model was applied for determination of silver species in personal care products. The errors oscillating up to 7.5% for the sum of Ag species are within the limits of the relative prediction error determined from the research studies while building the model.