Skip to main content

Created  by Mark Grujic.

Today, we are relaunching our web application to study Mineral Identification and Compositional Analysis (MICA).

Originally launched in 2020 as a submission to an app development competition, we find ourselves coming back to MICA as a demonstration of the kinds of visual storytelling that we try to do with complex datasets and ideas. Given an intricate dataset (it doesn’t have to be big), we want to identify key knowledge and relationships, and allow people to query and interact with their data in an intuitive and reliable way.

What is MICA?


MICA is a tool that aims to help geoscientists identify and group minerals that have similar chemical compositions. This is a fairly common problem, as the identification of a mineral based on its compositional chemistry is a difficult task. Hopefully MICA can make it a bit easier.

MICA is built from the database of minerals at which consists of 4722 minerals. The composition of 85 elements for each mineral is recorded in the database.

Through a process of dimensionality reduction and density-based clustering, we are able to present the local and global character of the variation of the 85 elements across all minerals in three dimensions, and group similar minerals into clusters, as seen above. The primary drivers for what makes each cluster unique are determined using a simple machine learning algorithm.


Image coloured by sulfer

At its core, MICA allows you to change the way you visualise minerals in relation to each other. Each point in the 3D visualisation represents a mineral. Similar minerals (as determined by elemental composition) are closer together, and far away from dissimilar minerals. The application Help tab has more information on distances. The colour and size of each represented mineral can be changed to focus on combinations of interest.

Given all this information and a query mineral that we want to further analyse, MICA provides the following information:

  • The most similar minerals (in terms of elemental composition) that could be confused

               ○ You can change the number of potentially confusing minerals

  • A list of all other minerals in the same cluster as the query

               ○ You can click on any of those minerals to set it as the target

  • A ranked list of elements that best separated the target mineral’s cluster from all others

The final piece of information that MICA provides is a holistic summary of elemental distribution within all 40 clusters. The Clusters tab shows the mean element composition, for all elements in all clusters. For example, cluster 13 has a mean Silicon proportion of 15.8%


And now in 2024…

When considering the techniques and tools used in MICA to solve the problem of mineral confusion, there have been substantial improvements over the last few years. In particular, the advances in graph-based network community detection and improved application development frameworks change how we address similar problems for our clients these days. 

That said, MICA is still used and useful, hosted and available to anybody on demand, and is able to be serviced with minimal overhead. At Datarock, we are on a journey to unearth possibility contained in our clients’ geoscientific datasets. We see building and hosting bespoke applications and solutions for our clients as a growing part of our future. We’ve learnt a lot in the last four years, culminating in our award winning cloud based drill core photography analysis platform. We are excited to share more details in this space in the coming months, so stay tuned!

Thanks for reading, and please have a play around with MICA. If you have any questions, please reach out.