MIST uses the International Mineralogical Association definitions and formulas for minerals. A mineral is a naturally occurring solid with well-defined chemical composition and crystallographic properties. MIST recognizes geochemical compositions that are consistent with a defined list of minerals based on geochemical relationships without structural information.

Stoichiometry is a field of chemistry based on ratios between chemical elements in reactions or minerals. MIST uses stoichiometry to search for mineral compositions in high resolution geochemical data. For example, if a composition contains one-to-one Na to Cl, MIST would report the mineral halite, or NaCl. If the composition were not balanced or contained too many other elements, that would indicate that there was another material mixed in and MIST would report a mixture.

Stoichiometry alone cannot perfectly define a mineral species, because minerals are defined by both chemistry and structure. However, geochemical observations are often easier to acquire on geological surfaces, and minerals, by definition, have specific proportions of elements bound within a crystal structure, which only allows some element substitutions. MIST incorporates the rule of the crystal structure requiring certain stoichiometric relationships to define a mineral AND the reality that many minerals allow some degree of element substitutions in real mineral structures to identify geochemical observations compositionally consistent with a single mineral crystal.

The model is agnostic to the source of the geochemical data acquisition method. The data should be quantitative enough to accurately represent ratios between chemical elements present. The key is having high resolution geochemical data.

Some common instruments and techniques are Electron Probe Microanalysis (EPMA), micro-X-Ray Fluorescence (XRF; e.g., PIXL instrument on the Mars 2020 rover Perseverance), Electron Dispersive Spectroscopy (EDS), Inductively coupled plasma mass spectrometry (ICP-MS), and Laser-Induced Breakdown Spectroscopy (LIBS; e.g., ChemCam and SuperCam instruments), among others.

This approach is ideal for datasets that have measurements smaller than the size of a grain, that is, where a measurement only takes into account a single grain. If data is from extremely fine grained material, you could expect to have many unidentified points or many points that are mixtures.

When minerals are similar in chemistry or structure, we rely on possible chemical substitutions to differentiate the minerals. Minerals have specific elements that can or cannot enter their crystal structure, even if they are in the same subgroup of minerals. Additionally, we implement a classification hierarchy from mineral class, to mineral group, to sub groups, to sub-sub groups, and finally to mineral species. This allows us to classify more general mineral groups before mineral species, and allows higher potential for identifications at some level of the hierarchy.

For example, it is tricky to separate a calcic amphibole and a clinopyroxene. These samples would first both be classified to the 'Silicate' mineral class and the 'Amph-Px' mineral group.

Then, we use >5 levels of checking the sum of cations, ratios of elements, and values of oxides to determine if the sample belongs in the 'Pyroxene' or 'Amphibole' subgroup.
In this example, it is important to check the amount of silicon (Si), calcium (Ca), potassium (K), and aluminum (Al) along with the sum of the cations in the A, B, C, D, and T positions of the crystallographic structure.

Next, the samples would go through >4 additional levels of checks within their respective subgroup (as described above) to determine which sub-sub group they belong in; calcic amphibole or clinopyroxene in this example.

Finally, a last layer of >3 checks will separate the samples in the calcic amphiboles or clinopyroxenes into their appropriate mineral species.

Importantly, this process is designed to identify any data compositionally consistent with a single mineral species.

This approach has limitations working with fine grained materials, when the grain is smaller than the measurement size. If a measurement contains geochemical data from multiple grains, this will lead to the algorithm identifying it as a mixture, or being unable to classify the data into any of our groups.

This approach cannot distinguish between mineral polymorphs.

This approach is not designed to take into account water that may be in a sample. This limits the distinctions between hydrated and non-hydrated minerals.

This model only identifies mineral groups and species that are coded into the model (see "What minerals can be identified" below). At this point, we have not incorporated reduced minerals like sulfides or metals, so those may be reported as mixtures or as corresponding oxidized minerals. In general, minerals not on the list would be interpreted as "mixtures."

MIST can only identify minerals that can be segregated using stoichiometry and that we have encoded into the algorithm*. When you run MIST, the output document will include the version number of MIST that was used in the calculation.

Click HERE to see a list of all minerals that can be identified by MIST!

*Please note that MIST will output the IMA mineral name and formula (some formulas are still a work in progress). If multiple mineral names share the same formula and have different structures or different hydration states, MIST may not list all possible mineral name options.

Option 1: Online

1. Go to the “Model” page.
2. Download the excel data input template, and carefully read the instructions on the first sheet.
3. Register your name, email, and affiliation. You will need to complete these fields each time you come back to the site (this is a work in progress). We will not share your data with anyone else.
4. Upload your input file, per instructions on the input template.
5. Run the model!
6. If the run completes successfully, clicking on the "Download Output" button will begin the download of your results. Information about the columns of the results file can also be found on the instructions page of the input template.
7. If no file downloads, please use the button to download the Error Log File. Email us at mist@mailman.rice.edu with the Error Log File and your input file for help.
8. Please contact us if you have any further problems or questions!

Option 2: MATLAB

Navigate to the GitHub repository to download the necessary files and run the code on your personal MATLAB.

Each column of the results document is outlined on the first sheet of the input template document (download on the "Model" page).

The results will preserve the input data, as well as provide the normalized data. Various variables important to the checks used in the model will be included in the table (such as calculation of specific cations or sum of cations). The results will also give the classification determined for each Group 1 (mineral class), Group 2 (mineral group), Group 3 (subgroup), Group 4(sub-sub group), and species. If a sample is identified as a species, there will also be a formula for that sample.

Yes! Visit our GitHub page to view and download all necessary files. Personal MATLAB lisence required.

The registration questions (name, email, affiliation) are for our eyes only. We will use this information to keep track of how MIST is being used, and how we could improve the model for our audience in the future.

Please email us at mist@mailman.rice.edu for any suggestions, corrections, errors, or additional information!

MIST interprets mineral crystals from their elemental components, so the MIST logo depicts a view into a mineral crystal where individual elements can be observed. The elements depicted form a silica tetrahedron, the primary building block of silicate minerals.

The logo was created by Andy Griffin.