SWIR Mineral Analysis with AI

Key Takeaways

SWIR mineral analysis uses short-wave infrared light (roughly 1,300 to 2,500 nm) to read mineral signatures that visible-light imaging cannot see, especially clay and alteration minerals.
Corescan’s Hyperspectral Core Imager and Malvern Panalytical’s ASD TerraSpec are the two dominant named instruments. One is built for continuous automated core scanning; the other is a portable spectrometer.
SWIR data increasingly feeds ML-based geomet and ore-sorting models, rather than only manual mineral identification by a geologist.
SWIR is a data-acquisition technology, not a decision-making AI on its own. The AI value comes from what you build on top: clay models, alteration domaining and sorting logic.

TL;DR

SWIR mineral analysis with AI starts by scanning core or ore with a hyperspectral instrument. Use Corescan’s HCI for continuous core logging, or a portable ASD TerraSpec for spot checks. Then a classifier turns the spectra into mineralogy that drives clay models and sorting logic.

What Is SWIR Mineral Analysis?

SWIR spectroscopy measures how a rock surface reflects light in the short-wave infrared band. Many minerals absorb light at specific wavelengths there. For example, bonds such as aluminium-hydroxyl, magnesium-hydroxyl and carbonate produce characteristic absorption features. Each clay or alteration mineral therefore leaves a distinctive fingerprint in the spectrum.

Those features are invisible to the naked eye and to standard RGB photography. That makes SWIR particularly valuable for identifying alteration mineralogy. Alteration is a first-order input for exploration targeting. It also matters for geometallurgy, because clays strongly affect flotation and comminution behaviour.

How Do I Analyze SWIR Mineral Data With AI?

The workflow has three stages: collect spectra, classify them, then use the results in models.

Collecting SWIR data from core and field samples

For continuous, automated coverage, Corescan’s Hyperspectral Core Imager scans drill core at roughly 800,000 spectral samples per metre. It covers the 450 to 2,500 nm range (VNIR through SWIR). It also co-registers the spectra with core photography and 3D laser profiling in a single pass. Corescan runs it as a service, through regional bureaus and on-site deployments, rather than as something you run entirely in-house. Our page on AI core scanning covers that workflow.

For spot checks or field use, Malvern Panalytical’s ASD TerraSpec is the established portable alternative. Geologists use it for fast pathfinder mineral identification in the field or core shed.

Classifying spectra into minerals

Raw spectral curves are not directly useful to a geologist or a downstream model. First, they need to become mineral species and abundance estimates. Automated spectral-matching algorithms do this job. Increasingly, ML classifiers trained on labeled spectral libraries do it too. As a result, SWIR mineral analysis can keep pace with thousands of metres of core.

Feeding SWIR mineral analysis into geomet and sorting models

Once you have classified mineralogy per metre of core, it becomes an input to hyperspectral clay models and alteration proxies. Those feed geometallurgical block models. SWIR is also becoming a real-time input to sensor-fusion ore sorting. Paired with XRT or optical sensors on systems like Steinert’s KSS, it adds mineralogical discrimination on top of density-based ore sorting sensors.

What to Expect From SWIR in Practice

SWIR reads only the surface it sees. Therefore, dusty, wet or weathered surfaces can distort results, and sample preparation matters. It also detects some minerals far better than others. Clays, micas, carbonates and sulphates respond well. By contrast, most sulphide ore minerals and quartz show no useful SWIR features.

Expect to validate classifications against XRD or automated mineralogy on a subset of samples. Finally, choose scale deliberately. Continuous scanning suits large drill programs, while a portable spectrometer suits smaller or targeted work.

FAQs

What minerals can SWIR mineral analysis detect?

SWIR works best on minerals with hydroxyl, water or carbonate bonds. That includes clays such as kaolinite and illite, micas, chlorite, carbonates and sulphates. It is weak for most sulphide ore minerals and for quartz.

How does AI help with SWIR data?

AI classifies raw spectra into mineral species and abundances automatically. It uses models trained on labeled spectral libraries. This turns huge volumes of scan data into mineralogy that geologists and geomet models can use.

Is SWIR the same as hyperspectral imaging?

Not exactly. SWIR is a wavelength band, while hyperspectral imaging describes capturing many narrow bands per pixel. Many hyperspectral core scanners, including Corescan’s HCI, cover SWIR alongside visible and near-infrared light.

Should I use a core scanner or a portable spectrometer for SWIR mineral analysis?

Use a core scanner like Corescan’s HCI for continuous coverage across large drill programs. Use a portable ASD TerraSpec for spot checks and field work. Many programs combine both.

Try It With Geocluster

Does continuous automated scanning (Corescan) or portable spot-checking (ASD TerraSpec) fit your program’s scale and budget? And how do you get the spectral data usefully into your geomet models? Geocluster can help you research both questions. You can also read how the Geocluster Research Harness works.