Times are displayed in (UTC-05:00) Central Time (US & Canada) Change

How Low Can You Go: Testing the Limits of LIBS Calibration Data with Respect to Laser Power, Texture, and Spectral Intensity

Session: Advancing Mineralogy and Spectroscopy Across the Solar System in Honor of MSA Roebling Medalist M. Darby Dyar

Presenting Author:

Jack Henry

Authors:

Henry, Jack D.¹, Siebach, Kirsten L.², Dyar, Melinda D.³, Lepore, Kate H.⁴, Ytsma, Cai R.⁵

(1) Department of Earth, Environmental, and Planetary Sciences, Rice University, Houston, TX, USA, (2) Department of Earth, Environmental, and Planetary Sciences, Rice University, Houston, TX, USA, (3) Department of Astronomy, Mount Holyoke College, South Hadley, MA, USA; Planetary Science Institute, Tucson, AZ, USA, (4) Department of Astronomy, Mount Holyoke College, South Hadley, MA, USA, (5) Cai Consulting, Glasgow, United Kingdom,

Abstract:

Laser-Induced Breakdown Spectroscopy (LIBS) uses a high-power laser to vaporize a target into plasma, allowing for rapid remote geochemical analyses. Chemistry is quantified from the emission spectrum of the plasma using multivariate machine learning models, trained on spectra of homogeneous samples with known compositions. Powdered standards are pressed into pellets and analyzed on the same instrument as the unknowns, creating a training dataset that captures the compositional variability anticipated in field data.

For LIBS applications on Mars, spectra are sometimes substantially different from those in the training set for various reasons (poor focus, poor coupling, mixed-phase targets, etc.). Such spectra present challenges for quantitative analysis. These spectra exhibit low total intensities, low signal to noise ratios, and are often collected on targets of rocks and loose soils. Ratios between different emission lines indicate low plasma temperatures that do not correspond with the spectral signals collected on calibration targets, making these spectra difficult to use for element predictions.

Previous work has shown that prediction uncertainty attributed to variations in LIBS emission intensity due to changing standoff distance can be mitigated using training data on the same pellets acquired over a range of laser powers. Here, we test whether a training dataset expanded to very low laser powers (0.4 – 1 mJ, over 3× lower than the lowest laser power previously analyzed) can model variations associated with analyzing loose powders finer than 38 μm, which produce low-temperature plasmas and have spectral intensities orders of magnitude lower than conventional training data.

By decreasing the laser power on pellet training data, meaningful spectra with lower plasma temperatures and low total signal intensity are produced, and compositions of pellets analyzed at these low laser powers can be modeled. However, these models still cannot accurately predict the compositions of loose fine powders, suggesting that issues other than plasma temperature are involved. Plasmas produced on loose powders likely incorporate atmospheric gas due to increased porosity and have distinct ionization peak ratios, possibly associated with geometric differences in the ablation pits or poor laser-sample coupling. Unfortunately, these differences are distinct enough to render the low laser power models inaccurate when predicting the powders. Ongoing work is dedicated to understanding and calibrating for these spectral effects and applying these expanded calibrations to Mars data.