This project is evaluating the use of changes in signals during single LA-ICP-MS analyses to identify inclusions within minerals, and the potential to improve the speed and consistency of LA-ICP-MS signal interpretation and processing. Each laser ablation ICP-MS point or line analysis contains an enormous amount of information, typically 30–40 elemental masses collected every 0.5 to 1 second during a single “sweep” over the mass spectrometer, over a 60-second interval. This means that every analysis yields between 1800 and 2500 data points. With hundreds of analyses collected on a daily basis, this data-rich environment presents an opportunity to apply new approaches including data analytics and machine learning to obtain more information from the data sets that the CODES Analytical Laboratories produce.

During 2019, our focus has been on developing workflows in open-source software packages in order to rapidly normalise LA-ICP-MS data signals, and to use machine learning approaches to automatically identify and classify LA-ICP-MS signals. In 2020, we plan to further develop and test these methods, and in particular focus on automatic mineral inclusion identification and classification.