Ultratrace analysis of Antarctic snow samples by reaction cell inductively coupled plasma mass spectrometry using a total-consumption micro-sample-introduction system

Abstract

In this work, a new sensitive procedure for the determination of ultratrace elements in snow samples based on quadrupole ICP-MS has been developed. After filtration through a 0.5 μm PTFE membrane (for dissolved element determination) or acidification with 0.5% nitric acid (for acid dissolvable element determination), the analytes were preconcentrated by sample volume reduction and analysed by ICP-MS. Micro-samples were efficiently introduced into the plasma source at 20 μl min⁻¹ uptake rate by using a PFA micronebulizer coupled to an evaporation chamber of the torch integrated sample introduction system (TISIS). As a result, the amount of sample required was about one order of magnitude lower than that required with a conventional liquid sample introduction system. In order to improve the transport efficiency, the TISIS chamber was electrically-heated at 70 °C and a sheathing gas stream was used to protect the aerosol from the chamber walls. Under these conditions, negative solvent plasma effects were no more severe than for conventional systems, because the total solvent plasma load was 20 mg min⁻¹. The operating parameters were optimized to obtain maximum sensitivity, while limiting oxides and double charge ion formation. The polyatomic interferences were removed by applying the dynamic reaction cell (DRC) technique, using ammonia as the reaction gas. Under the optimized conditions, limits of detection ranged from 0.02 to 4.5 pg g⁻¹, allowing the determination of Cr, V, Fe, Mn, Pb, Zn, Cd, Co and Cu in Antarctic snow samples. Signal repeatability was lower than 10% which prevented the use of an internal standard. Precision of the procedure ranged from 2.0% to 5.6%. The accuracy of the method was verified by the analysis of both certified reference water and surface snow samples collected in coastal and inland areas of Antarctica. The DRC program used, the short wash out and signal stabilization times registered under these conditions led to a 10 h⁻¹ sample throughput.