Improving quantitation of nanoparticle concentrations in mixtures of known components using single particle inductively coupled plasma mass spectrometry with kernel density estimation

Abstract

Accurate quantification of nanoparticle fractions in polydisperse mixture is challenging for many analytical techniques. For single particle inductively coupled plasma mass spectrometry (spICP-MS) where pulse signals are correlated to nanoparticles size and number concentration, data processing is crucial. The state-of-the-art algorithms distinguish particulate components with size difference of more than about 20 nm. In this work, a Bayes discriminant analysis based on a finite mixture model consisting of kernel density estimation (FMKDE) has been applied for spICP-MS, resulting in an unprecedented improvement in size resolution. Compared to existing data processing methods, the FMKDE model provides highly accurate contents due to the flexibility of the kernel density estimation for probability density. The gold nanoparticle mixtures have been used to verify the comparison. The size discrimination limit for gold mixtures can be decreased to about 5 nm with resolvable component content of <10% using the FMKDE model. The quantified number concentrations are also consistent with the designed values for the prepared mixtures with absolute biases of <3%. This study improves the size resolution of spICP-MS which could broaden its applicability in particle mixtures.