| Résumé | Using a calibration model built from 14 compressed powders of reference materials doped with gold in solution, the gold concentration of 32, 6-sided, ore samples was determined by LIBS and then compared to those obtained by conventional laboratory techniques (gravimetry/atomic absorption). The studied samples are on average ~ 4 cm long and they were collected from 6 different gold mines of the Abitibi greenstone belt (Canada). LIBS measurements are found to be accurate for homogeneous powder samples but can differ significantly from laboratory measurements for solid ore samples depending on the uniformity of the gold distribution at the sampling scale due to the diferent sampling approaches adopted by both techniques, i.e. surface sampling versus bulk. In addition, the LIBS sampling strategy, which consists of scanning along diagonals oriented transversely to the natural ore layers, was evaluated on the same 32 samples. This sampling strategy results in a gold concentration very close to that obtained by scanning the entire rock surface, but with only a small fraction (of the order of the square root) of the number of laser shots. Further, the influence of the chemical composition of the calibration samples on the properties of the plasma was studied by measuring the electron number density and temperature of the plasma. These parameters were found to be almost identical (1.70 ± 0.15 × 10¹⁶ cm⁻³ and 6200 ± 1000 K) within the uncertainty of the measurements despite the large difference in iron content (3% to 50%) of the samples. |
|---|
