Three different kinds of serpentine mineral samples were investigated using Fourier transform near-infrared spectroscopy (FTNIR). The results show that there are obvious differences in the characteristic infrared spectra of the three serpentine group minerals (lizardite, chrysotile, and antigorite), which can easily be used to identify these serpentine minerals. The combination of weak and strong peaks in the spectrum of lizardite appears at 3650 and 3690 cm⁻¹, while the intensities of the peaks at 4281 and 4301 cm⁻¹ (at 7233 and 7241 cm⁻¹, respectively) are similar. A combination of weak and strong peaks in chrysotile appears at 3648 and 3689 cm⁻¹ and at 4279 and 4302 cm⁻¹, and a single strong peak appears at 7233 cm⁻¹. In antigorite, there are strong single peaks at 3674, 4301, and 7231 cm⁻¹, and the remaining peaks are shoulder peaks or are not obvious. The structural OH mainly appears as characteristic peaks in four regions, 500–720, 3600–3750, 4000–4600, and 7000–7600 cm⁻¹, corresponding to the OH bending vibration, the OH stretching vibration, the OH secondary combination vibration, and the OH overtone vibration, respectively. In the combined frequency vibration region, the characteristic peak near 4300 cm⁻¹ is formed by the combination of the internal and external stretching vibrations and bending vibrations of the structural OH group. The overtone vibrations of the OH stretching vibration appear near 7200 cm⁻¹, and the practical factor is about 1.965. The near-infrared spectra of serpentine minerals are closely related to their structural differences and isomorphous substitutions. Therefore, near-infrared spectroscopy can be used to identify serpentine species and provides a basis for studies on the genesis and metallogenic environment of these minerals.

The identification of minerals is indispensable in geological analysis. Traditional mineral identification methods are highly dependent on professional knowledge and specialized equipment which often consume a lot of labor. To solve this problem, some researchers use machine learning algorithms to quickly identify a single mineral in images. However, in the natural environment, minerals often exist in an associated form, which makes the identification impossible with traditional machine learning algorithms. For the identification of associated minerals, this paper proposes a deep learning model based on the transformer and multi-label image classification. The model uses transformer architecture to model mineral images and outputs the probability of the existence of various minerals in an image. The experiments on 36 common minerals show that the model can achieve a mean average precision of 85.26%. The visualization of the class activation mapping indicates that our model can roughly locate the identified minerals.

Pyrophyllite and talc are both tetrahedra–octahedra–tetrahedra (TOT)-type phyllosilicates, but differences can be found in the stacking mode of the layers and the ion occupation. Fourier transform infrared spectroscopy was used to differentiate between pyrophyllite and talc. In the 400–600 cm⁻¹ region, pyrophyllite exhibits six peaks, while talc only exhibits five peaks. In the 1000–1200 cm⁻¹ region, pyrophyllite exhibits three clear peaks at approximately 1051, 1070, and 1121 cm⁻¹; while talc only exhibits one strong peak near 1020 cm⁻¹. The differences between pyrophyllite and talc in the near–infrared (NIR) region are clear in the 4000–4700 cm⁻¹ region, and pyrophyllite exhibits an intense peak around 4615 cm⁻¹, which is attributed to the combination of the OH and Si-O-Si stretching bands. Talc has a maximum peak located near 4324 cm⁻¹, which is attributed to the OH stretching vibration. In addition, talc has a secondary peak near 4366 cm⁻¹. Talc has two other weaker peaks around 4054 and 4180 cm⁻¹. The 7000–7250 cm⁻¹ region exhibits the first fundamental overtone of the OH group stretching vibrations. The common characteristic band of these two minerals is 7175–7183 cm⁻¹. The first overtone of OH stretching vibrations can also be generated by adjacent peaks in the fundamental overtones. The peaks of these two minerals around 7094 cm⁻¹ appear to be a combination of 3630 (±5) cm⁻¹ and 3642 (±3) cm⁻¹.The factor of the first fundamental overtone of the OH group stretching vibration is 1.95 (±0.003). Therefore, the characteristic peaks in the mid-infrared (MIR) and NIR regions can be used to distinguish between pyrophyllite and talc, providing a research basis for further exploration in related geological areas.

Kaolinite and dickite are differently ordered polytypes of kaolinite-group minerals, whose differences are in the stacking mode of layers and ion occupation. Fourier transform infrared spectroscopy was used to collect information about the differences between the two minerals. The common characteristics of kaolinite and dickite are bands near 4530 and 7068 cm⁻¹, which are attributed to the combination of the inner Al-OH stretching vibration and outer Al-OH bending vibration and the overtone of the inner Al-OH stretching vibration, respectively. The difference is that kaolinite has secondary peaks at 4610 and 7177 cm⁻¹, and the secondary peak of dickite is near 4588 cm⁻¹. The OH stretching vibration has the first fundamental overtone of the stretching vibration in the range of 7000–7250 cm⁻¹. In addition to the overtones generated by single OH stretching vibrations, overtones combining different OH stretching vibrations are also found, which are formed by adjacent peaks of OH stretching vibrations. The average factor of the first fundamental overtone with an OH-group stretching vibration is approximately 1.95. The near-infrared spectrum (NIR) of phyllosilicates is closely related to their structure and isomorphism. Therefore, the near-infrared region can distinguish between kaolinite and dickite and provide a basis for deposit research and geological remote sensing.

A rare massive yellowish-green serpentinized dunite, covering a minimum area up to ~50 m², has been found in Ji’an County, Jilin Province, Northeast China. It contains primary olivine and secondary serpentine (antigorite) and brucite. Other primary minerals like orthopyroxene, clinopyroxene, and aluminum-rich phase (such as garnet, spinel, and plagioclase), frequently appearing in ultramafic rocks, have not been identified. The olivine is essentially pure forsterite, with an Mg# (100 × Mg/(Mg + Fe)) of ~99.6–99.7. Due to these distinct features, we especially name the protolith of this dunite as jianite (集安岩). The forsterite grains range up to ~2 mm, show clear equilibrium textures such as nearly straight grain boundaries and ~120° dihedral angles at their triple junctions, and display no intragranular or intergranular composition variations. They are extensively ruptured and hydrated (i.e., serpentinized), with the fractures (and the grain boundaries as well) filled by fine-grained antigorite (ideally Mg₆(Si₄O₁₀)(OH)₈) and brucite (ideally Mg(OH)₂). These secondary phases are also extremely poor in Fe, indicating a good chemical equilibrium with the forsterite. The serpentinization reaction may have proceeded as forsterite + fluid = antigorite + brucite at temperatures of ~425(25) °C and at relatively low but undetermined pressures. The fluid was likely a B-rich, but Si-poor dilute aqueous fluid, as implied by the trace element characteristics and water-related infrared features of the forsterites in equilibrium. The petrogenesis of the jianite is presently unclear.

Mineral recognition is of importance in geological research. Traditional mineral recognition methods need professional knowledge or special equipment, are susceptible to human experience, and are inconvenient to carry in some conditions such as in the wild. The development of computer vision provides a possibility for convenient, fast, and intelligent mineral recognition. Recently, several mineral recognition methods based on images using a neural network have been proposed for this aim. However, these methods do not exploit features extracted from the backbone network or available information of the samples in the mineral dataset sufficiently, resulting in low recognition accuracy. In this paper, a method based on feature fusion and online hard sample mining is proposed to improve recognition accuracy by using only mineral photo images. This method first fuses multi-resolution features extracted from ResNet-50 to obtain comprehensive information of mineral photos, and then proposes the weighted top-k loss to emphasize the learning of hard samples. Based on a dataset consisting of 14,986 images of 22 common minerals, the proposed method with 10-fold cross-validation achieves a Top1 accuracy of 88.01% on the validation image set, surpassing those of Inception-v3 and EfficientNet-B0 by a margin of 1.88% and 1.29%, respectively, which demonstrates the good prospect of the proposed method for convenient and reliable mineral recognition using mineral photos only.