ULTRATRACE IMPURITIES DETERMINATION IN SILVER FOR NUCLEAR APPLICATIONS: SIGNAL OPTIMIZATION IN ICP-MS WITH ARTIFICIAL NEURAL NETWORKS AND RESPONSE SURFACE METHODOLOGY

SILVANA MARTIN; DAMIÁN PIUSELLI; JORGE MAGALLANES; DIEGO ONNA; EZEQUIEL MORZAN

Simposio; 15° Rio Symposium on Atomic Spectrometry; 2019

Productionof alloys for the nuclear application requires analyzing impuritiesin metals used as raw material. The specifications for nuclear gradesilver-indium-cadmium alloys define maximum levels of impurities oneach metal. Therefore, several impurities must be quantified in eachindividual metal. Recently,the ICP-MS technique has acquired great relevance for impuritiesquantification since it allows multi-elemental analysis at theultra-trace level. The quality of the measurement depends on thesignal optimization that maximizes the sensitivity and leads to lowerlimits of detection. To improve the performance, operationalvariables must be optimized and, in general, the "one variableat a time" (OVAT) optimization has been used. However, thismethodology is not the most efficient to obtain the maximum responsevalues1.Thiswork presents an empirical model using a full factorial design forthe ICP-MS analytical signal optimization of several elements basedon three variables: the torch position, the makeup gas flow, and thenebulizer flow. Response surface methodology and artificial neuralnetworks (ANN) were compared to describe the empirical model and abetter description was obtained with the ANN. The model was testedanalyzing certified reference material and spiked samples in theoptimal conditions, finding no significant bias. The figures of meritof the optimized method and the results of the analyzed silversamples will be shown.Inconclusion, the empirical model created by ANN helps to understandthe influence of operational variables on the analytical signal. Theoptimized method helped quantify ultra-trace impurities in a complexmatrix with maximum sensitivity.p { margin-bottom: 0.25cm; direction: ltr; line-height: 115%; text-align: left; orphans: 2; widows: 2 }p.western { font-size: 11pt; so-language: pt-BR }p.cjk { font-size: 11pt; so-language: ja-JP }p.ctl { font-size: 11pt }a:link { color: #0563c1 }a.ctl:link { font-family: "Times New Roman" }