Skip to content
Publicly Available Published by De Gruyter September 30, 2016

Automated Quantitative Rare Earth Elements Mineralogy by Scanning Electron Microscopy

  • Sven Sindern EMAIL logo and F. Michael Meyer
From the journal Physical Sciences Reviews

Abstract

Increasing industrial demand of rare earth elements (REEs) stems from the central role they play for advanced technologies and the accelerating move away from carbon-based fuels. However, REE production is often hampered by the chemical, mineralogical as well as textural complexity of the ores with a need for better understanding of their salient properties. This is not only essential for in-depth genetic interpretations but also for a robust assessment of ore quality and economic viability. The design of energy and cost-efficient processing of REE ores depends heavily on information about REE element deportment that can be made available employing automated quantitative process mineralogy.

Quantitative mineralogy assigns numeric values to compositional and textural properties of mineral matter. Scanning electron microscopy (SEM) combined with a suitable software package for acquisition of backscatter electron and X-ray signals, phase assignment and image analysis is one of the most efficient tools for quantitative mineralogy. The four different SEM-based automated quantitative mineralogy systems, i.e. FEI QEMSCAN and MLA, Tescan TIMA and Zeiss Mineralogic Mining, which are commercially available, are briefly characterized.

Using examples of quantitative REE mineralogy, this chapter illustrates capabilities and limitations of automated SEM-based systems. Chemical variability of REE minerals and analytical uncertainty can reduce performance of phase assignment. This is shown for the REE phases parisite and synchysite. In another example from a monazite REE deposit, the quantitative mineralogical parameters surface roughness and mineral association derived from image analysis are applied for automated discrimination of apatite formed in a breakdown reaction of monazite and apatite formed by metamorphism prior to monazite breakdown.

SEM-based automated mineralogy fulfils all requirements for characterization of complex unconventional REE ores that will become increasingly important for supply of REEs in the future.

1 Introduction

The Rare Earth Elements (REEs) comprise the lanthanides, a group of 15 metals, as well as Sc and Y because they tend to occur in the same mineral assemblage as the lanthanides and exhibit similar chemical properties. All the rare earths except Pm are available in the geological environment. Following a common pattern within the periodic table, the lanthanides with even atomic numbers are more common in nature, and in general, REE with lower atomic numbers are more common ionic constituents in REE minerals and occur in greater abundance than the REE with higher atomic numbers. The trends in the Earth’s crustal abundance suggest a division between light and heavy REEs [1]. Although variation exists, commonly the lanthanides are subdivided according to their atomic weight into light rare earths (LREEs) comprising La through Gd, and heavy rare earths (HREEs) including Tb through Lu, and Y.

Industrial consumption of REE indicates that demand is driven by the rapid development of emerging industries. These include hybrid electric vehicles and wind power generation, which rely on the use of rare earths such as Nd for permanent magnets. Other consumer sectors include FCC catalysts, battery alloys, metallurgical products, phosphors, ceramics, catalytic converters, as well as chemical, military and satellite industries [2].

Given this global demand for green and sustainable products in energy, military and manufacturing industries, REE demand throughout the world is projected to increase. Consequently, the REEs are increasingly becoming more attractive commodity targets for the mineral industry. Almost all current mine production of REEs is derived from less than ten minerals, and in particular, bastnaesite, monazite and xenotime [3]. There is, however, a problem especially with monazite as it may contain deleterious impurities like Th, which imparts an unwanted radioactivity to the ores. The cost of handling and disposing of radioactive material is a serious impediment to the economic extraction of the more radioactive REE-rich minerals [4]. Another problem arises from the fact that the REE portfolio produced by current mining is dominated by Ce and La, accounting for 77 % of the total world supply. There is no indication of change in the near future since known ore reserves of producing mines have an even higher share of Ce plus La with close to 80 %. This leads to an imbalance between the REE mix produced by current mining and the industrial demand, as not all of the 15 REEs are of the same industrial importance. For example, REEs used for high-performance permanent magnets are Pr, Nd, Tb and Dy, while Eu, Tb, Er and Y are components of other products such as phosphors and ceramics [5]. The imbalance between industrial need for certain REEs and the mine production portfolio can be assessed by the critical REE ratio of (La + Ce)/(Nd + Eu + Tb + Dy + Y). This ratio is 4.5 for the 2014 production and 5.4 for the REE reserves known in the same year. Adding up the 2014 commodity prices [6] for each REE, this ratio is 0.00545. In other words, the cumulative unit value for Nd + Eu + Tb + Dy + Y is 183.5 times higher than that for La + Ce.

This unfavourable ratio calls for a paradigm shift from conventional high LREE ores to unconventional REE ores with HREE enrichments. The distribution of individual REEs in minerals is not even and certain minerals will hold enrichments or depletions of light or heavy REEs, respectively. This renders most REE ores mineralogically and chemically complex. This, together with a wide textural variability, demands sufficient understanding of the salient characteristics of REE ores, essential for a robust assessment of the quality and economic viability of a potential REE deposit. In that sense, it is important to identify and quantify all the REE-phases present in the ore and to evaluate the elemental deportment of the different REEs into the different REE-bearing phases.

Following the production chain, mineral processing requires additional information on mineral associations, grain size distributions as well as grain liberation characteristics. Nowadays, mineral processing is generally capable of sequential separation of multiple mineral phases, but when the REEs are found in two or more mineral hosts, with each behaving differently during extraction, beneficiation can become relatively inefficient and costly [7]. From that, it would appear that deposits in which the REEs are preferentially concentrated in a single mineral phase have a competitive advantage. On the other hand, mining of REEs as by-product of additional or major commodities has a significant price advantage since extraction costs are shared by the bundle of metals mined. However, polymetallic ores, by nature, contain a complex and diverse mineral assemblage, which may require special beneficiation strategies to become technically and economically viable. The design of cost-efficient processing for such ores depends heavily on information that can be made available employing automated quantitative process mineralogy [8].

2 Quantitative mineralogy

Quantitative mineralogy assigns numeric values to compositional and textural properties of mineral matter (e.g. rocks, ores and processing products). Compositional properties are related to the kind and quantity of mineral phases in a sample. This is known as the modal composition, which specifies the relative proportions in volume% or mass% from highly abundant rock forming minerals to less abundant accessory phases.

Textural properties comprise a multitude of information on the grains forming the sample as well as on their spatial arrangement. Grains are characterized by their size and shape. Further quantitative mineralogical parameters related to size are grain size distributions, either of the total sample or a specific mineral fraction. The spatial distribution of grains can be described by properties such as alignment, orientation or spacing [9]. Quantitative mineralogy thus attempts to convert rock properties to numbers, which often are only described in a qualitative way by geoscientists. As numeric procedures can only handle quantitative data, this represents an important contribution to the development of computer models, e.g. on beneficiation processes [10].

A variety of methods including polarization microscopy, electron microscopy, X-ray diffraction, differential thermal analysis and infrared and optical emission spectrometry has been applied in the last decades for quantitative mineralogical studies, e.g. [1113].

Automated procedures to generate quantitative mineralogical data require unattended identification of mineral phases. Furthermore, not only to account for target elements of mining but also in order to identify deleterious elements (e.g. U, Th), such systems should allow element detection. Any of the previously mentioned textural properties can only be quantified if information on phases and chemistry are related to images. Imaging must be performed with high spatial resolution to be suitable for fine-grained rocks. Thus, accessory phases or complex intergrowth of minerals can reliably be quantified. Finally, all data sets must be available in digital form for image analysis. Scanning electron microscopy (SEM) meets most of these requirements and is one of the most efficient tools for quantitative mineralogy.

3 Scanning electron microscopy

In a SEM, an electron beam is generated from a tungsten filament or a field emission gun. The electron beam is focussed and can be directed on a single spot or it can be scanned over a specimen. Among the interactions between the electrons penetrating the specimen and the sample elements, the electrons scattered back from the incident beam (backscatter electrons, BSEs) and fluorescence X-ray photons generated by excitation of atoms in the specimen are most important for quantitative mineralogy. Both yield information on sample composition (Figure 1).

Figure 1: Schematic sketch showing interaction volume of electrons in sample, not to scale, according to Refs [14, 16, 17]. Pear shape is typical of materials with low mean atomic numbers such as silicate or carbonate minerals.
Figure 1:

Schematic sketch showing interaction volume of electrons in sample, not to scale, according to Refs [14, 16, 17]. Pear shape is typical of materials with low mean atomic numbers such as silicate or carbonate minerals.

The intensity of the BSE signal is primarily correlated to the mean atomic number of the sample [14]. This allows differentiation of minerals; in particular, minerals containing elements with higher atomic number, such as the REE, are characterized by high BSE intensity and can be distinguished from silicates or carbonates composed of lighter elements. The energy of fluorescence photons is characteristic of the excited elements while their intensity is proportional to the concentration of these elements. As both BSE and fluorescence photons are generated by an interaction of sample atoms and electrons, they represent a larger volume of the specimen than only the area of the incident electron beam ([15], Figure 1). In an untitled sample, the volume of interaction depends on mean atomic number, density and beam energy [16, 17]. The latter is a function of the potential applied for acceleration of electrons emitted from the electron gun. At usual analytical conditions, i.e. acceleration voltages between 15 and 25 kV, the width of the interaction volume limits the spatial resolution of the technique in most mineral matrices to values between 2 and 5 μm.

Polished samples – either as a thin section on glass or a polished epoxy resin mount – are preferred to avoid surface effects on BSE intensity. As most mineral samples are not conductive, the samples are coated with carbon prior to analysis.

4 SEM-based automated quantitative mineralogy

Software packages, which make a conventional SEM applicable for automated quantitative mineralogy, are complex key components of such analytical systems. They control the entire analytical procedure of one or more samples in an unattended way. This comprises control and optimization of instrumental parameters as well as acquisition and storage of BSE and X-ray signals. Furthermore, software packages must offer a diversity of image analysis functions.

Grains with coherent properties and their boundaries must be identified in an image built up by the signals according to the scanning pattern (segmentation, e.g. [18, 19]). In a further step, the grains must be mineralogically identified, and the software should be able to separate touching grains for further image analysis (de-agglomeration). Data are transformed to digital images displaying mineral species with user-defined colours. Proportions of different minerals (i.e. modal abundance) can be calculated simply by counting pixels quantifying the area assigned to each mineral. According to the Delesse principle, this is equivalent to volume proportions. Using density information that is included in a database implemented in the software package modal abundance in volume% can be transferred to mass%.

These basic digital image data allow computation of quantitative textural parameters. Grains, either imaged in a mount of separate grains or numerically de-agglomerated from an image of a rock sample, can further be characterized in terms of size parameters (area, perimeter and diameter). Many software packages derive these values after calculation of equal area circles or ellipses for each grain. Furthermore, grains can be characterized with respect to their shape applying mathematical shape parameters. Numerous parameters converting quantitative values of grain-radius, {-diameter,} -area or -perimeter to grain shape are described in the literature (i.e. aspect ratio [20], circularity [21], convexity [22], eccentricity [23], projection shape factor [24], projection sphericity [25], radial shape factor, roughness factor [24], solidity [22] and surface roughness [23]).

Image data can also be used for the quantification of textural parameters that, for example, describe orientation as well as contact relationships or mineral associations.

Orientation can be expressed as a statistics on the angles of long axes of elongated particles. Pixels along shared boundaries of two phases can be counted yielding a statistics of mineral associations if applied to all mineral grains forming contacts in a rock sample.

Finally, quantitative image analysis data can be combined to derive parameters that are directly applicable to processing of ores and to the evaluation of beneficiation processes. Mineral liberation serves as a good example. It may be either defined as the modal proportion of an ore mineral in a particle or as the proportion of an ore mineral boundary, which is associated with the mounting resin to the total boundary of a particle. The latter indicates the free surface of an ore mineral amenable to processing operations (e.g. flotation, hydrometallurgy).

According to the authors’ knowledge, currently, there are four SEM-based automated quantitative mineralogy systems commercially available (in alphabetical order of companies): FEI (QEMSCAN, MLA), Tescan (TIMA) and Zeiss (Zeiss Mineralogic Mining).

4.1 Quantitative Evaluation of Minerals by Scanning Electron Microscopy

Early generations of the QEMSCAN (Quantitative Evaluation of Minerals by Scanning Electron Microscopy) were developed by CSIRO (Commonwealth Scientific and Industrial Research Organisation) in Australia in the middle of the 1970s as QEM*SEM automated mineragraphy [15, 26]. The actual system is based upon FEI Quanta 650 and Quanta 650F SEMs, and in addition to a standard version, it is offered as product models “express” and “well site” (www.fei.com). The instrument is equipped with a Bruker Dual X-Flash5030 energy-dispersive X-ray spectroscopy (EDS) detector. Up to 14 round epoxy-mounted samples (diameter 30 mm) or 12 standard thin sections can be placed in the sample chamber.

The workflow of a QEMSCAN analysis is displayed in Figure 2. The specimen is scanned on a grid basis with an electron beam step width usually varying between 0.8 and 10 μm (Figure 2(a) and 2(b)). For each step, which is represented by an image pixel in the resulting dataset, the system acquires the BSE intensity and X-ray fluorescence spectrum consisting of 1,000–5,000 counts at acceleration voltages of 15, 20 or 25 kV (Figure 2(c)). Information on up to 72 elements is derived from peak signals within the spectrum. Count rates at spectral peaks are transferred to information on element abundance.

Figure 2: Workflow of a QEMSCAN Analysis, (a) Pattern of frames on a carbon-coatedpolished thin section of a rock. Frames are sequentially scanned. (b) Each frame isscanned with an electron beam step width varying between 0.8 μm (minimumvalue) and usually 10 μm. (c) Acquisition of BSE intensity and fluorescence spectrumwith EDS X-ray detectors usually on the basis of 1,000–5,000 counts. (d) Datastorage on hard disc and phase assignment to each pixel by correlation ofX-ray and BSE data with a database. (e) Generation of images with colour coding fordifferent mineral phases. (f) Image analysis, generation of quantitative mineralogicaldata (e.g. modal composition, particle size and size distribution, mineral association,particle shape and textural parameters).
Figure 2:

Workflow of a QEMSCAN Analysis, (a) Pattern of frames on a carbon-coatedpolished thin section of a rock. Frames are sequentially scanned. (b) Each frame isscanned with an electron beam step width varying between 0.8 μm (minimumvalue) and usually 10 μm. (c) Acquisition of BSE intensity and fluorescence spectrumwith EDS X-ray detectors usually on the basis of 1,000–5,000 counts. (d) Datastorage on hard disc and phase assignment to each pixel by correlation ofX-ray and BSE data with a database. (e) Generation of images with colour coding fordifferent mineral phases. (f) Image analysis, generation of quantitative mineralogicaldata (e.g. modal composition, particle size and size distribution, mineral association,particle shape and textural parameters).

Mineral identification is performed comparing pixel information with a database (Species Identification Protocol, SIP) defining a range of element compositions, BSE values and detector count rates for each phase. Mineral assignment is done at first match within a sequence of phases also defined in the SIP (Figure 2(d) and 2(e)). Image data generated in these steps constitute the basis for further quantitative mineralogical evaluation of the samples (Figure 2(f)).

Among the measurement modes of QEMSCAN, two are most often applied. The fieldscan/field image mode is recommended for the study of non-fragmented rocks showing their original texture. This mode generates images displaying the total scanned area, which offers data for mineral abundance and textural analysis. An example of such a fieldscan image is given in Figure 3, which shows a rock formed by a groundmass of plagioclase and muscovite as well as by larger irregularly shaped quartz grains indicated by different colours. Less abundant mineral phases, such as the REE-mineral monazite (REEPO4) as well as apatite or clinozoisite-epidote, are detected by QEMSCAN and can easily be recognized due to dark blue, dusky pink and lavender colour coding, respectively.

Figure 3: QEMSCAN fieldscan image of a monazite (REEPO4$\text{REEPO}_{4}$) bearing rock. Minerals are indicated by colour coding as indicated in the image (apatite = dusky pink, chlorite = dark green, Fe-oxide = brown, monazite = dark blue, muscovite = light green, plagioclase = bluish green, quartz = pink and zoisite-epidote = lavender). The sample areas of insets a, b and c are alternatively displayed as BSE image on the right side. Monazite is characterized by a high density compared to other phases and thus shows highest BSE intensity. Contours of three quartz grains are highlighted by red lines in inset a. BSE scaling appropriate for distinction of minerals with elevated density (e.g. apatite, Fe-oxide and monazite) in insets b and c is not suitable for distinction of lighter phases like muscovite, plagioclase and quartz, which dominate in inset a.
Figure 3:

QEMSCAN fieldscan image of a monazite (REEPO4) bearing rock. Minerals are indicated by colour coding as indicated in the image (apatite = dusky pink, chlorite = dark green, Fe-oxide = brown, monazite = dark blue, muscovite = light green, plagioclase = bluish green, quartz = pink and zoisite-epidote = lavender). The sample areas of insets a, b and c are alternatively displayed as BSE image on the right side. Monazite is characterized by a high density compared to other phases and thus shows highest BSE intensity. Contours of three quartz grains are highlighted by red lines in inset a. BSE scaling appropriate for distinction of minerals with elevated density (e.g. apatite, Fe-oxide and monazite) in insets b and c is not suitable for distinction of lighter phases like muscovite, plagioclase and quartz, which dominate in inset a.

The time-consuming effort of high-resolution pixel-by-pixel X-ray mapping is highly appropriate for the study of materials composed of fine-grained minerals with similar BSE properties. This is illustrated in Figure 3. While phase assignment for muscovite, plagioclase and quartz is clear using X-ray data, these minerals, which are characterized by similar and relatively low densities, can hardly be distinguished in BSE images at a BSE scale suitable for distinction of heavier phases.

In contrast to the fieldscan mode for non-fragmented rocks, samples consisting of particulate material (crushed rocks, ore concentrates and unconsolidated sediments) mounted in a resin representing specimen surface that is not of interest are best analysed with the “Particle Mineral Analysis” (PMA) mode. Here, sample areas not exceeding a user-defined BSE threshold value indicative of background resin are not analysed. Further, measurement modes of QEMSCAN are described in Ref. [26].

4.2 Mineral Liberation Analyser

Originally, the MLA was developed by the JKMRC (Julius Kruttschnitt Mineral Research Centre, University of Queensland) in the 1990s [27]. Today, the system is provided by the FEI Company and the hardware basis is identical to QEMSCAN.

Its software also allows pixel-by-pixel determination of BSE and EDS data along a grid pattern on the sample (GXMAP mode). The characteristic and probably most often used measurement mode discriminates grains first on the basis of their BSE intensity, which is calibrated to a range of atomic numbers. Finally, a single X-ray analysis is performed on each grain for phase assignment in combination with the BSE values according to best spectral match of analytical data with those of mineral standard spectra available in a mineral reference library [19]. This XBSE mode reduces analysis time due to limitation of the time-consuming step of EDS analysis. It is most appropriate for samples composed of larger grains with coherent BSE values. Further, measurement modes are described in Refs [19, 27, 28].

4.3 Tescan-Integrated Mineral Analyser

The hardware basis of the Tescan-integrated mineral analyser (TIMA) is either formed by a MIRA field emission or a VEGA thermal emission SEM equipped with up to four EDS detectors (www.tescan.com). BSE and spectral information obtained for each pixel is used for identification of mineral phases. So far, literature information on this recently released analytical system is not available.

4.4 ZEISS Mineralogic Mining

The Mineralogic Mining System was released by Zeiss in 2014. It can be run on the different SEM platforms offered by the company (e.g. Zeiss EVO, Zeiss SIGMA HD, as well as mobile systems designed for on-site mineral processing purposes, MinScan, www.zeiss.com). The software allows various modes of data acquisition [29]. Mineral classification can be performed using BSE greyscale values as well as in combination with X-ray spectra that are acquired by one or more EDS detectors for all image pixels. Furthermore, in a mode comparable to MLA, touching pixels with common BSE intensity are combined and mineral identification is performed using a single EDS analysis.

Prior to phase assignment, X-ray data are transferred to quantitative element concentrations [29, 30]. Element wt.% values are then compared to minimum and maximum ranges indicative for phases of interest listed in a database. The use of quantitative values (i.e. data directly related to mineral stoichiometry) makes phase assignment independent of instrumental parameters affecting the EDS spectrum (e.g. acceleration voltage [29]).

To enhance the volume of information and to support phase identification, the system allows the correlation of SEM data with images obtained from external sources (e.g. polarization light and cathodoluminescence microscopy, 2D virtual slices through 3D X-ray tomographic datasets [30]).

5 Quantitative REE mineralogy

REE minerals are chemically highly variable and show complex formulae. Rocks in which they occur in high abundance that could serve as REE ores generally show complicated textures. Both, petrologists who want to understand the origin of such rocks and mining engineers who develop concepts for ore processing need to study chemical and textural variability of such minerals. This represents a challenge for automated quantitative mineralogy but also a field for which this technology is highly applicable, as will be illustrated by following examples.

Owing to mainly identical charges (3+) and minor variations in ionic radii, various REEs isostructurally substitute in mineral phases. Most REE minerals are thus characterized by a range of chemical compositions rather than by one specific stoichiometry.

For example, the common REE ore minerals parisite and synchysite, which can in a simplified way be expressed as CaREE2(CO3)3(F, OH)2 and CaREE(CO3)2(F, OH), respectively, do not only show incorporation of the different lanthanoids La, Ce, Pr, Nd or Sm but also replacement of these elements by Y and Th as well as of Ca by other elements such as Sr, Ba or Fe. A review of published data [3134] on these minerals reveals the following compositional ranges for both minerals (in part recalculated for Ca + Sr + Ba + Fe = 1 per 3 CO3):

  • Parisite: Ca0.891.00Sr0.000.04Ba0.000.09Fe0.000.03La0.290.67Ce0.771.16Pr0.080.11Nd0.180.37Sm0.010.05Eu0.000.01Gd0.000.02Y0.000.04Th0.000.01(CO3)3(F, OH)2

  • Synchysite: Ca0.891.00Sr0.000.03Fe0.000.03La0.240.39Ce0.470.53Pr0.030.06Nd0.080.19Sm0.010.02(CO3)2(F, OH)

Calculated fluorescence spectra representative of these variable compositions obtained in an SEM analysis are displayed in Figure 4. In accordance with the mineral stoichiometry, parisite is characterized by higher signals of REE as well as F and lower signals of Ca compared to synchysite. However, the signal ranges of La, Ce, Pr and Nd for both minerals show an overlap. Furthermore, automated quantitative mineralogical systems dominantly use spectral information from limited photon counts (e.g. 1,000–5,000) for phase assignment – independent of the software solutions offered by different companies. Thus, counting statistics, detector noise as well as surface effects lead to uncertainty of signal intensity. In consequence, chemical variability of REE minerals as illustrated by the example in Figure 4 and analytical uncertainty can reduce the performance of phase assignment. In such cases, minerals that cannot be distinguished have to be combined in a group of phases.

Figure 4: Fluorescence spectra of synchysite and parisite calculated for mineral compositions published by [31–34], calculation performed with the iDiscover® software by FEI-company assuming an acceleration voltage of 15 kV. The indices K and L denote spectral information belonging to the {K- and} L-series, respectively. Only in the case of Ca Kα$K_{{{\alpha}}}$ (Ka) and Kβ$K_{{{\beta}}}$ (Kb), radiation is indicated.
Figure 4:

Fluorescence spectra of synchysite and parisite calculated for mineral compositions published by [3134], calculation performed with the iDiscover® software by FEI-company assuming an acceleration voltage of 15 kV. The indices K and L denote spectral information belonging to the {K- and} L-series, respectively. Only in the case of Ca Kα (Ka) and Kβ (Kb), radiation is indicated.

Monazite (REEPO4) is one of the most important ore minerals for LREE. It is an important constituent of magmatic as well as of placer deposits and belongs to the first mineral phases used for extraction of REE [1].

In addition to a quantitative evaluation of a REE ore, which implies the determination of the modal abundance of monazite or other REE minerals, it is important to understand ore-forming processes. The definition of parameters controlling monazite stability that are indicated by monazite-forming as well as by monazite-consuming mineral reactions is an important basis for such study. This is illustrated in Figure 5, which displays details of the QEMSCAN image in Figure 3 showing a monazite-bearing metamorphic rock.

Figure 5: Example of image analysis applied for the distinction and quantification of primary apatite and apatite formed by breakdown of monazite. (a) Grains 1–4 are characterized by the parameters association and surface roughness. Association indicates the proportion of monazite pixels in contact with an apatite grain in % of the total amount of pixels defining the perimeter of the apatite grain. The surface roughness is defined as (2×π×(area pixels/π)1/2$2\times \pi \times (\textrm{area pixels}/\pi)^{1 / 2}$/perimeter pixels). Apatite formed by reaction of monazite (grains 1 and 2) is characterized by high association and low surface roughness values. (b) Position of apatite grains 1 and 2 in a reaction rim around monazite, sample area is identical to inset b in Figure 3(c)). Examples of primary metamorphic apatite grains (3 and 4) in a matrix composed of plagioclase and muscovite with minor chlorite, sample area identical to inset c in Figure 3.
Figure 5:

Example of image analysis applied for the distinction and quantification of primary apatite and apatite formed by breakdown of monazite. (a) Grains 1–4 are characterized by the parameters association and surface roughness. Association indicates the proportion of monazite pixels in contact with an apatite grain in % of the total amount of pixels defining the perimeter of the apatite grain. The surface roughness is defined as (2×π×(area pixels/π)1/2/perimeter pixels). Apatite formed by reaction of monazite (grains 1 and 2) is characterized by high association and low surface roughness values. (b) Position of apatite grains 1 and 2 in a reaction rim around monazite, sample area is identical to inset b in Figure 3(c)). Examples of primary metamorphic apatite grains (3 and 4) in a matrix composed of plagioclase and muscovite with minor chlorite, sample area identical to inset c in Figure 3.

Monazite (REEPO4) shown by dark blue colour coding is surrounded by a corona of apatite (Ca5(PO4)3(OH, F) (dusky pink), which formed in a breakdown reaction indicating the limit of monazite stability. The reaction can be formulated in a simplified way as:

3REEPO4(monazite)+5Cain fluid2++(F,OH)in fluidCa5(PO4)3(F,OH)apatite+3REEin fluid3+.

The abundance of apatite thus could be considered as an indicator for this breakdown reaction, which was probably triggered by a retrograde albitization of plagioclase releasing Ca2+ to the fluid. However, the image shows that next to fine-grained and irregularly shaped apatite from the coronas, a second type can be observed. Apatite of this type, which is texturally not associated to monazite and which forms larger rounded grains with smooth boundaries, is genetically not related to monazite. Rather, it belongs to the original metamorphic assemblage.

Both types of apatite must be distinguished in order to indicate apatite as monazite breakdown product. Shape parameters allow transferring the descriptive characterization of both types to measurable numeric data in an automated quantitative mineralogical approach that is independent of the experience or optical impression of a mineralogist performing individual microscopic grains inspection.

The surface roughness parameter compares the actual perimeter of a grain (formed by neighbouring pixels of the same kind) with the perimeter of a circle with equivalent area quantified by the amount of pixels representing the grain. Thus, the regularly shaped apatite grains with smooth boundaries, which belong to the metamorphic assemblage, are characterized by high roughness parameters close to 1, whereas the irregularly shaped apatite grains forming the coronas around unstable monazite yield low roughness parameters.

Grain association is another parameter of image analysis that can be applied for discrimination of both types, because apatite formed by monazite breakdown generally is in contact with monazite. In contrast, metamorphic apatite is almost exclusively associated to other minerals than monazite.

A combination of both parameters, roughness and association, thus provides a criterion for automated discrimination of both types of apatite, which is illustrated in Figure 5. In a series of samples, image analysis could be applied to quantify volumes of both reaction partners and consequently to show progress of monazite breakdown.

6 Concluding remarks

Although based on concepts from the 1970s automated quantitative mineralogy is a field that has shown significant progress in the last years due to hardware and software developments of the SEM manufacturers. The examples outlined in this chapter underline the capabilities of automated quantitative mineralogy for phase identification and textural analysis. Therefore, this technology fulfils all requirements for characterization of complex unconventional REE ores with HREE enrichment that will become increasingly important for supply of REEs in the future.

Acknowledgements

This article is also available in: Golloch, Handbook of Rare Earth Elements. De Gruyter (2016), isbn 978–3–11–036523–8.

QEMSCAN measurements by R. Klinghardt and L. Gronen are gratefully acknowledged. We thank G. Günther for retrieval of literature.

References

[1] Chakmouradian AR, Wall F. Rare earth elements: Minerals, mines, magnets (and more). Elements 2012, 8, 333–40.10.2113/gselements.8.5.333Search in Google Scholar

[2] Alonso E, Sherman AM, Wallington TJ, Everson MP, Field FR, Roth R, Kirchain RE. Evaluating rare earth element availability: A case with revolutionary demand from clean technologies: Environ Sci Technol 2012, 46, 3406–14.10.1021/es203518dSearch in Google Scholar

[3] Hellman PL, Duncan RK. Evaluation of rare earth element deposits: IMM Trans B: Appl Earth Sci, 2014,123, 107–17.10.1179/1743275814Y.0000000054Search in Google Scholar

[4] USEPA. Rare earth elements: A review of production, processing, recycling, and associated environmental issues: Cincinnati, U.S. Environmental Protection Agency (USEPA), 2012, 135.Search in Google Scholar

[5] USDOE. Critical mineral strategies: US Department of Energy, 2010, 165.Search in Google Scholar

[6] HEFA, Rare earth metal prices: HEFA Rare Earth Inc., www.mineralprices.com. 2014. Accessed: October 15, 2015.Search in Google Scholar

[7] Jordens A, Cheng YP, Waters KE. A review of the beneficiation of rare earth element bearing minerals. Miner Eng 2013, 41, 97–114.10.1016/j.mineng.2012.10.017Search in Google Scholar

[8] Gu Y, Schouwstra RP, Rule C. The value of automated mineralogy. Miner Eng 2014, 58, 100–3.10.1016/j.mineng.2014.01.020Search in Google Scholar

[9] Butcher AR. Automated mineralogy & petrography for geometallurgy. 11th Freiberg short course in economic geology 2012, December 5–7, 175p.Search in Google Scholar

[10] Lund C, Lamberg P. Geometallurgy – A tool for better resource efficiency. Eur Geol 2014, 37, 39–43.Search in Google Scholar

[11] Lyon RJ, Tuddenham WM. Quantitative mineralogy as a guide in exploration. Min Eng 1959, 11, 1–5.Search in Google Scholar

[12] Davies BL, Walawender MJ. Quantitative mineralogical analysis of granitoid rocks: A comparison of X-ray and optical techniques. Am Miner 1982, 67, 1135–43.Search in Google Scholar

[13] Kile DE, Eberl DD. Quantitative mineralogy and particle-size distribution of bed sediments in the Boulder Creek Watershed, Colorado, in: Murphy SF, Verplanck PL, Barber LB (eds.) Comprehensive water quality of the Boulder Creek Watershed, Colorado, during high-flow and low-flow conditions, 2000. US Geological Survey Water-Resources Investigations Report 03–4045, 2003, 173–84.Search in Google Scholar

[14] Knight RD, Klassen RA, Hunt P. Mineralogy of fine-grained sediment by energy-dispersive spectrometry (EDS) image analysis – a methodology. Environ Geol 2002, 42, 32–40.10.1007/s00254-002-0538-7Search in Google Scholar

[15] Pignolet-Brandom S, Reid KJ. Mineralogical characterization by QEM*SEM, in: Carson DJ, Vassiliou AH (eds.) Process Mineralogy VIII. Warrendale, PA, USA, The Minerals, Metals & Materials Society 1988, 337–46.Search in Google Scholar

[16] Potts PJ. A Handbook of Silicate Rock Analysis. London, Glasgow, Weinheim, New York, Tokyo, Melbourne, Madras, Blackie Academic & Professional, 622p.Search in Google Scholar

[17] Jones MP. Methoden der Mineralogie. Stuttgart, Ferdinand Enke Verlag, 260p.Search in Google Scholar

[18] Heilbronner R, Barret S. Image Analysis in Earth Sciences. Berlin, Heidelberg, Springer-Verlag, 2014.10.1007/978-3-642-10343-8Search in Google Scholar

[19] Gu, Y. Automated scanning electron microscope based mineral liberation analysis. J Miner Met Charact Eng, 2003, 2, 33–41.10.4236/jmmce.2003.21003Search in Google Scholar

[20] Schneiderhöhn P. Eine vergleichende Studie über Methoden zur quantitativen Bestimmung von Abrundung und Form an Sandkörnern. Heidelb Beitr Mineral Petrogr 1954, 4, 172–91.10.1007/BF01111154Search in Google Scholar

[21] Rodriguez J, Johansson J, Edeskär T. Particle shape determination by two-dimensional analysis in geotechnical engineering. Proceedings of Nordic Conference on Soil Mechanics and Geotechnical engineering, NGM 2012, 207–18.Search in Google Scholar

[22] Olson E. Particle shape factors and their use in image analysis – part 1: Theory. J GXP Compliance 2011, 15, 85–96.Search in Google Scholar

[23] Podczeck F, Newton JM. The evaluation of a three-dimensional shape factor for the quantitative assessment of the sphericity and surface roughness of pellets. Int J Pharm 1995, 124, 253–59.10.1016/0378-5173(95)00096-2Search in Google Scholar

[24] Bouwman AM, Bosma JC, Vonk P, Wesselingh, JH, Frijlink HW. Which shape factor(s) best describe granules? Powder Technol 2004, 146, 66–72.10.1016/j.powtec.2004.04.044Search in Google Scholar

[25] Podczeck F, Rahman SR, Newton JM. Evaluation of a standardised procedure to assess the shape of pellets using image analysis. Int J Pharm 1999, 192, 123–38.10.1016/S0378-5173(99)00302-6Search in Google Scholar

[26] Pirrie D, Rollinson GK. Unlocking the applications of automated mineral analysis. Geol Today 2011, 27, 235–44.10.1111/j.1365-2451.2011.00818.xSearch in Google Scholar

[27] Sylvester P. Use of the Mineral Liberation Analyzer (MLA) for mineralogical studies of sediments and sedimentary rocks, in: Sylvester P (ed.) Quantitative Mineralogy and Microanalysis of Sediments and Sedimentary Rocks. Mineralogical Association of Canada Short Course Series 2012, 42, 1–16.Search in Google Scholar

[28] Fandrich R, Gu Y, Burrows D, Moeller K. Modern SEM-based mineral liberation analysis. Int J Miner Process 2007, 84, 310–20.10.1016/j.minpro.2006.07.018Search in Google Scholar

[29] Graham SD, Brough C, Cropp A. An introduction to Zeiss Mineralogic Mining and the correlation of light microscopy with automated mineralogy: a case study using BMS and PGM analysis of samples from a PGE-bearing chromitite prospect. Precious Metals’15, 2015, Falmouth UK, 10p.Search in Google Scholar

[30] Graham SD, Cropp A. Advances in mineral characterisation techniques: emerging tools and workflows. Proceedings of the 13th Biennial SGA Meeting 2015, 24–27 August, Nancy France, 1403–5.Search in Google Scholar

[31] Manfredi TR, Bastos Neto AC, Pereira VP, Barbanson L, Schuck C. The parisite-(Ce) mineralization associated with the Fazenda Varela carbonatite (Correia Pinto, SC). Pesquisas em Geosciencas 2013, 40, 295–307.10.22456/1807-9806.77831Search in Google Scholar

[32] Petrella L. The nature and origin of REE mineralization in the Misery syenitic intrusion, northern Québec, Canada. Master thesis 2012, McGill University Montreal, 130p.Search in Google Scholar

[33] Ni Y, Post JE, Hughes JM. The crystal structure of parisite-(Ce), Ce2CaF2(CO3). Am Miner 2000, 85, 251–8.10.2138/am-2000-0126Search in Google Scholar

[34] Zaitsev AN, Wall F, LeBas MJ. REE-Sr-Ba minerals from the Khibina carbonatites, Kola Peninsula, Russia: their mineralogy, paragenesis and evolution. Miner Mag 1998, 62, 225–250.10.1180/002646198547594Search in Google Scholar

Published Online: 2016-9-30
Published in Print: 2016-9-30

© 2016 by Walter de Gruyter Berlin/Boston

Downloaded on 28.3.2024 from https://www.degruyter.com/document/doi/10.1515/psr-2016-0063/html
Scroll to top button