Reflectance spectroscopy & hyperspectral imaging

Spectral reflectance data is captured by illuminating a material with a light source and recording the reflected light with a spectrometer or hyperspectral line scan camera. The collected data can provide information on the mineralogy because certain minerals absorb light at specific wavelengths. The processes through which these absorptions take place vary between the different regions of the electromagnetic spectrum. Only a limited set of minerals can be detected on each of these regions because the absorption processes require a specific molecular configuration of the mineral. The table below presents an overview of how well different types of minerals can be identified with spectral reflectance measurements on the VIS-NIR, SWIR and LWIR spectral ranges. The MWIR spectral range is not included in this table because reflection spectroscopy applications for minerals is relatively unexplored on this spectral range.

Spectral reflectance sensors are useful tools for mapping the geology of ore deposits^2,3. This is especially the case for hydrothermal deposit types. At these deposits, ore minerals originate from hot watery fluids that are produced by some sort of volcanic activity. Ore minerals are deposited from the fluids because the pressure and temperature changes when it travels to the surface and fractures the rocks. This mineralisation is often associated with chemical alteration of the minerals that were originally in the rocks. Some infrared reflectance sensors are extremely sensitive to the chemical changes within these minerals and are therefore useful for identifying different mineralisation zones and structures. Even though many ore minerals can not be detected with spectral reflectance sensors, detection of these altered minerals can often be used to distinguish different ore types and ranges of ore grade (waste/low/high). The table below presents an overview of different environments of ore deposit formation and the characteristic mineral assemblages that can be detected with spectral reflectance sensors.

For applications in hydrothermal deposit mining, an additional calibration step of the reflectance sensors is needed to link the detection of altered minerals to ore grade. This requires measurement of a set of samples that covers the mineralogical variability of the deposit. A large number of samples (> 1000) is often needed in order to use AI and machine learning to unravel any indirect relationships. Drill core scanning is considered to be useful for this because drill core samples are often already collected with the aim of covering geological variability. This can also serve two purposes since the drill core scans can be used to improve deposit models and simultaneously reveal opportunities for scanning applications during mine operation.

The physical processes through which absorptions take place differ between the different spectral regions. The table below shows an overview of these processes for each region.

The absorption mechanism that is most often observed in the VIS-NIR spectra of minerals is the crystal field effect⁴. Crystal field effects are absorptions of light by electrons that change between the orbital energy levels of transition metal complexes. Transition metal complexes are molecules consisting of a transition metal and a ligand. See this article on Libretexts chemistry for more info.

The wavelength of crystal field absorptions is related to the energy difference between the orbitals. This energy difference mainly depends on the type of transition metal and ligand that form the complex⁴. This means that the wavelength of absorption depends on the mineral in which the transition metal occurs. Transition metals for which absorptions due to crystal field effects are often observed include Fe, Ni, Cu, Mn, Cr and Ti⁵.

Charge transfer is the migration of electrons between neighbouring ions or between ions and ligands^4,5. This can also occur between neighbouring ions of transition metals in different valence states, such as Fe²⁺ and Fe³⁺. The wavelength of absorption depends on the energy that is required for the charge transfer. Charge transfer absorptions are commonly seen in the reflectance spectra of iron oxides. As an example, the figure below presents the VIS-NIR spectrum of the iron-oxide mineral goethite. The reflectance drop-off towards 500 nm in this spectrum is produced by charge transfer absorption. The absorption features that occur around 650 and 900 nm are caused by electronic transitions.

Another type of absorption that is produced by certain minerals are colour centres. As with crystal field effects, colour centres are also absorptions produced by electrons that change between energy levels. However, in this case this results from a defect in the crystal structure of the mineral. For more information on crystals and crystallography, see this article on Libretexts chemistry.

Two types of colour centres can be distinguished. The first type is due to a vacancy defect in which an electron can reside at different discrete energy levels^5,6. This is referred to as an electron colour centre. An example of a mineral for which an electron colour centre is often observed is fluorite. The other type of colour centre is referred to as a hole colour centre and is produced when an electron is missing from a location where there is normally an electron pair⁶. This can be caused by the presence of an impurity ion. A hole colour centre is for example responsible for the colour of smoky quartz.

Absorptions from conduction band transitions can often be observed in the VIS-NIR spectra of semiconductor minerals, such as many types of sulphides⁷. In these minerals electrons can be attached to individual atoms, known as the valence band, or they can move freely through the lattice, known as the conduction band^4,5. The energy difference between these two bands is known as the band gap. Electrons can transfer from the valence band to the conduction band by absorbing light with energies larger than the band gap⁵. This produces a step function in the VIS-NIR spectra of these minerals. The wavelength location of the step depends on the mineral in which the conduction band transition takes place. Examples of minerals for which such absorptions are observed include realgar, stibnite, and native sulphur⁵. An example of a VIS-NIR spectrum of realgar is presented below.

Molecule bonds vibrate by stretching of the bond or by bending of the angle of the bond. In polyatomic molecules, multiple bonds vibrate at the same time and produce different types of group vibrations⁹. This article on Libretexts chemistry shows an overview of the different group vibrations.

The molecule bond vibrations can also be excited with two or more energy quanta, producing so-called overtones of the fundamental vibrational modes. Additionally, two or more fundamentals and/or overtones of different modes can add or subtract to form so-called combinations^4,9. Minerals and mineral characteristics can be identified from the wavelength position and intensity of a specific combination of absorption features resulting from the fundamental, overtone and combinations of molecule bond vibrations. The figure below presents an example of a measured MWIR spectrum on a quartz sample. It clearly shows the absorption features from molecule bond vibrations as minima on the spectral profile.

The occurrence of water also produces absorption features in the SWIR. This water can be physically adsorbed on the surface of mineral grains, it can occupy specific lattice sites or it can be part of the crystal structure¹². The figure below presents an example of the SWIR spectra of several minerals on which the wavelength ranges at which the most diagnostic molecule bond vibrations are indicated. Since not all molecule bonds are diagnostic in the SWIR spectral region, the use of SWIR spectrometry is generally limited to the following groups of minerals; phyllosilicates, hydroxylated silicates, sulphates, carbonates and ammonium-bearing minerals¹¹.

The figure below presents an example of the MWIR-LWIR diffuse reflectance spectrum of a quartz sample. The reflectance peaks around 8-10 µm and 12-13 µm are the reststrahlen features of the Si–O stretching vibrations of quartz.

In the MWIR-LWIR region of the electromagnetic spectrum materials also emit black-body radiation. The intensity of the radiation that is emitted is related to the temperature of the material by Planck’s law. Temperature differences can influence the MWIR-LWIR reflectance spectra that are measured.

The spot size of point spectrometers is usually around several square centimetres. However, any spot size is theoretically possible by configuring the collection optics of the measurement system.

Using a line scan sensor to collect hyperspectral imagery data is currently possible up to around 800 fps on the VIS-NIR range, 450 fps on the SWIR range, 370 fps on the MWIR range, and 240 fps on the LWIR range^15,16. However, the frame rate is also related to spectral resolution, spatial resolution and the signal-to-noise ratio. Customization of the optical collection system is possible to obtain a higher frame rates with lower resolution or signal-to-noise for example. Additionally, it should be noted that hyperspectral imagery data on the MWIR and LWIR range is usually of much lower quality in terms of signal-to-noise than point spectra obtained with a Fourier-transform spectrometer.

Reflectance measurements require a light source to illuminate the materials. In earth observation spectroscopy and large area scans, sunlight is usually the only available source. However, the intensity and spectral profile of the sunlight is not constant, which means corrections are needed during the data analysis that may affect measurement accuracy. In applications such as drill core and conveyor belt scanning, halogen lamps are most often used for illumination.

Absorption features on the SWIR and MWIR range are more narrow and distinct than those on the VIS-NIR. However, the features of different minerals often still overlap and limit the detection capabilities of minerals occurring in low quantities.

The reststrahlen features of different minerals in the LWIR have lots of overlap. Unmixing these features into different minerals is therefore difficult. This is easier if the spot or pixel size of the reflectance measurements is small in relation to the crystal or grain size of minerals in the rock because pure mineral components are more frequently measured in that case.

Detection limits only concern minerals with distinct absorption features that occur in a mineral mixture. The detection limit in this case is the surface area coverage at which the mineral’s features can still be observed in the measured spectrum.

The detection limit mainly depends on the overall reflectivity of a rock or mineral. Detection limits are higher on dark rocks because less light is reflected, which reduces the signal-to-noise of the recorded spectrum. A higher amount of noise conceals the mineral absorption features at lower concentrations.

If multiple minerals are measured, the spectrum is a combination of the absorption profiles of each mineral. Certain minerals produce more intense absorption features than others, and the absorption intensity is not linearly related to concentration in terms of area coverage. The intensity of a mineral’s absorption profile can also change if it is in a 50-50 mixture and the other mineral is replaced by a different one. This is especially the case if the other mineral was white and transparent, and is replaced by a dark and opaque one.

The chapter on “spectroscopy” within the open-access library of Libretexts chemistry also provides a good introduction that can be accessed for free.

• USGS spectral library
• CSIRO mineral spectral libraries
• RRUFF database

Additionally, the ECOSTRESS spectral library can be freely accessed to view mineral spectra, but payment is needed to download data. The number of views per day is also restricted.

Mineral reference spectra that can be used to analyse reflectance sensor measurements are also published in literature. Publications that are considered to be particularly useful are the following:

• “Visible and near infrared spectra of minerals and rocks” by Hunt & Salisbury (1970, Modern geology)
• “High spectral resolution reflectance spectroscopy of minerals” by Clark et al. (1990, Journal of geophysical research).
• “G-MEX spectral interpretation field manual” by AusSpec (2008, AusSpec international ltd.)

Finally, commercial software is available for automated analysis of mineral reflectance spectra. Mineral identification with such software is usually based on an integrated spectral library. Some of the libraries mentioned above are usually included with the software.