Laser-induced breakdown spectroscopy; LIBS; Variable angle; Incidence angle; Collection angle; Partial least-squares analysis; PLS;

Laser-induced breakdown spectroscopy has become a popular tool for rapid elemental analysis of geological materials. However, quantitative applications of LIBS are plagued by variability in collected spectra that cannot be attributed to differences in geochemical composition. Even under ideal laboratory conditions, variability in LIBS spectra creates a host of difficulties for quantitative analysis. This is only exacerbated during field work, when both the laser-sample distance and the angle of ablation/collection are constantly changing. A primary goal of this study is to use empirical evidence to provide a more accurate assessment of uncertainty in LIBS-derived element predictions. We hope to provide practical guidance regarding the angles of ablation and collection that can be tolerated without substantially increasing prediction uncertainty beyond that which already exists under ideal laboratory conditions. Spectra were collected from ten geochemically diverse samples at angles of ablation and collection ranging from 0 degrees to +/- 60 degrees. Ablation and collection angles were changed independently and simultaneously in order to isolate spectral changes caused by differences in ablation angle from those due to differences in collection angle. Most of the variability in atomic and continuum spectra is attributed to changes in the ablation angle, rather than the collection angle. At higher angles, the irradiance of the laser beam is lower and produces smaller, possibly less dense plasmas. Simultaneous changes in the collection angle do not appear to affect the collected spectra, possibly because smaller plasmas are still within the viewing area of the collection optics, even though this area is reduced at higher collection angles. A key observation is that changes in the magnitude of atomic and total emission are <5% and 10%, respectively, in spectra collected with the configuration that most closely resembles field measurements (W) at angles <20 degrees. In addition, variability in atomic and continuum emission is strongly dependent upon sample composition. Denser, more Fe/Mg-rich rocks exhibited much less variability with changes in ablation and collection angles than Si-rich felsic rocks. Elemental compositions of our variable angle data that were predicted using a much larger but conventionally-collected calibration suite show that accuracy generally suffers when the incidence and collection angles are high. Prediction accuracy (for measurements acquired with varying collection and ablation angles) varies from +/- 1.28-1.86 wt% for Al2O3, +/- 1.25-1.66% wt for CaO, +/- 1.90-2.21 wt% for Fe2O3T, +/- 0.76-0.94 wt% for K2O, +/- 2.85-3.61 wt% MgO, +/- 0.15-0.17 wt% for MnO, +/- 0.68-0.78 wt% for Na2O, +/- 0.33-0.42 wt% for TiO2, and +/- 2.94-4.34 wt% SiO2. The ChemCam team is using lab data acquired under normal incidence and collection angles to predict the compositions of Mars targets at varying angles. Thus, the increased errors noted in this study for high incidence angle measurements are likely similar to additional, unacknowledged errors on ChemCam results for non-normal targets analyzed on Mars. Optimal quantitative analysis of LIBS spectra must include some knowledge of the angle of ablation and collection so the approximate increase in uncertainty introduced by a departure from normal angles can be accurately reported. (C) 2017 Elsevier B.V. All rights reserved.