The new ArDI web solution makes it easier to identify tiny micro-inclusions in minerals without destroying the sample, accelerating geological exploration and discovery of new deposits.
A team of scientists from several institutes of the Russian Academy of Sciences presented the ArDI (Advanced spectRa Deconvolution Instrument) web application, which radically simplifies the identification of microcrystals in fluid and melt inclusions of minerals. The application analyzes spectra obtained by Raman spectroscopy (RS) and allows you to determine the composition of the sample. To handle complex inclusions, ArDI combines neural networks, Faiss index search and spectrum decomposition algorithms. This makes it much easier to identify micron-sized minerals, even if they are hidden within other crystals.
Inclusions in minerals are microscopic capsules that store gases and melts – valuable sources of information about the formation of ores and the evolution of magma. Diagnostics of such objects is difficult to carry out using traditional methods: most tools require destruction of the sample or are not suitable for very small sizes. Raman spectroscopy eliminates this barrier, providing reliable data on the structure and composition without destruction. However, the task of spectral recognition remains nontrivial due to signal overlap and variability in the properties of minerals. That’s why scientists created ArDI, a hybrid system that speeds up processing and increases recognition accuracy to 96%.
The ArDI platform is available online and supports searching across several databases, including RRUFF and ROD. Its own database includes about 3800 reference spectra of more than 1000 substances. The developers note that when loading a spectrum, information about the mineral being studied and its source must be indicated, which ensures the quality and accuracy of identification. According to the creators, ArDI already saves time and resources on analysis, and also has the potential for use in interplanetary exploration and space missions, where spectral methods remain one of the main tools.
The development of ArDI will continue by expanding the database and introducing new algorithms for processing spectra, which will make solving mineralogical problems more efficient. According to the authors, ArDI is ahead of many similar solutions in terms of the number of reference spectra and recognition accuracy, second only to RRUFF.
Source: IGM SB RAS
