LIBS-Acoustic Mid-Level Fusion Scheme for Mineral Differentiation under Terrestrial and Martian Atmospheric Conditions

Abstract

The shockwave produced alongside the plasma during a laser-induced breakdown spectroscopy event can be recorded as an acoustic pressure wave to obtain information related to the physical traits of the inspected sample. In the present work, a mid-level fusion approach is developed using simultaneously recorded laser-induced breakdown spectroscopy (LIBS) and acoustic data to enhance the discrimination capabilities of different iron-based and calcium-based mineral phases, which exhibit nearly identical spectral features. To do so, the mid-level data fusion approach is applied concatenating the principal components analysis (PCA)-LIBS score values with the acoustic wave peak-to-peak amplitude and with the intraposition signal change, represented as the slope of the acoustic signal amplitude with respect to the laser shot. The discrimination hit rate of the mineral phases is obtained using linear discriminant analysis. Owing to the increasing interest for in situ applications of LIBS + acoustics information, samples are inspected in a remote experimental configuration and under two different atmospheric traits, Earth and Mars-like conditions, to validate the approach. Particularities conditioning the response of both strategies under each atmosphere are discussed to provide insight to better exploit the complex phenomena resulting in the collected signals. Results reported herein demonstrate for the first time that the characteristic sample input in the laser-produced acoustic wave can be used for the creation of a statistical descriptor to synergistically improve the capabilities of LIBS of differentiation of rocks.

Figure 1

(a) Panoramic view of the TVC in the UMALASERLAB. (b) Experimental setup scheme. (c) Optical setup placed inside the TVC: 1, Laser mirror and telescope’s secondary mirror; 2, Telescope’s primary mirror; 3, Optical fiber holder; 4, Microphone without its hemi-anechoic chamber; 5, Temperature monitoring thermocouple. (d) Sample holder inside the hemi-anechoic chamber.

Figure 2

Data fusion process scheme.

Figure 3

Average LIBS spectra for the analyzed samples remarking the presence of some detected impurities. (a) Average spectra of Fe-rich minerals, (b) spectra (265–290 nm) showing emission lines of Fe II-red dashed/shadow, Mg I and II-blue, and Si I-green dashed. (c) Spectra (375–400 nm) with Fe I-red shadow, Ca II-purple dashed, and Al II-bright blue. (d) Average spectra of Ca-rich minerals. (e) Spectra (260–290 nm) with Mg I and II. (f) Spectra (415–465 nm) with Ca I-red dashed/shadow and Sr I and II-orange dashed, and Mg II.

Figure 4

PCA scores for the first three PCs. Cold tones correspond to Fe-rich materials, while warm tones correspond to the Ca-rich subgroup.

Figure 5

(a) Average time domain signals in the 0.1–1.25 ms range. (b) Average amplitude in the first maximum of the acoustic wave for each mineral studied. (c) Slope value obtained for each mineral. Values closer to 0 indicated stable signal during the laser shot series, while negative values implied decreases in the measured sound intensity.

Figure 6

Graphical representation of the new LIBS-acoustic descriptor for Fe-rich minerals (left) and Ca-rich minerals (right). A factor is applied to each element to improve the visualization.

Figure 7

Percentage confusion matrix for the studied samples.

Figure 8

PCA scores for the first three PCs. Cold tones correspond to Fe-rich materials, while warm tones correspond to the Ca-rich subgroup.

Figure 9

Percentage confusion matrix for the studied samples measured in Mars-like atmosphere.