Datasets of measurements for cores A, B and C.

Laser-induced breakdown spectroscopy (LIBS) appears to be a promising technique for rapid on-site assessment of precious metal concentrations in ores. However, a number of issues need to be considered for the optimal use of this technique in practical situations. This article focuses on the number of measurements (i.e., spectra or laser shots) required to assess the mean palladium concentration in drill cores from the Lac des Iles mine (Ontario, Canada). We have performed a probabilistic study of the accuracy of the mean palladium concentration obtained by LIBS as a function of the number of measurements at random locations. For this purpose, we first used the results of a detailed laser scan of the core surface and then a mathematical model of the probability density of the palladium distribution to explore the parameter space, in particular the effect of noise on the measurements. We show that for the 1-meter-long, 2-centimeter radius quarter core samples analyzed, a few thousand randomly sampled locations generally provide an assessment of the palladium concentration within useful confidence limits. For a typical laser repetition rate of 50 Hz, such an analysis is a matter of minutes compared to hours or days using conventional methods.