Unsupervised and interpretable discrimination of lithium-bearing minerals with Raman spectroscopy imaging

Area of Science:

• Geoscience
• Materials Science
• Data Science

Background:

• Lithium-bearing minerals are vital for energy storage and advanced technologies.
• Conventional mineral identification methods are often subjective, costly, or time-consuming.
• Raman Spectroscopy (RS) offers detailed molecular information for accurate mineral characterization.

Purpose of the Study:

• To develop an unsupervised methodology for lithium-mineral identification using Raman Imaging.
• To address the need for efficient and accurate tools in lithium resource exploration and processing.
• To differentiate minerals with similar elemental compositions, such as petalite and spodumene.

Main Methods:

• Development of an unsupervised machine-learning solution for Raman Imaging data.
• Utilizing specific spectral bands characteristic of lithium-bearing minerals.
• Testing the methodology's robustness with blind samples.

Main Results:

• The machine-learning approach accurately identifies and differentiates lithium minerals.
• The method provides interpretable results based on distinct spectral signatures.
• Robustness confirmed through successful analysis of blind samples.

Conclusions:

• Unsupervised Raman Imaging with machine learning offers a reliable and efficient method for lithium-mineral identification.
• This technique enhances accuracy in mineral differentiation, crucial for the lithium-mining industry.
• The study provides insights into spectral features enabling precise mineral analysis.