This application note describes the use of XRD, together with the Rietveld method and cluster analysis, for phase identification and quantification of heavy mineral sands.
XRD can be used to readily identify the mineral phases present in heavy mineral sands, such as ilmenite FeTiO3, rutile TiO2, zircon ZrSiO4, quartz SiO2, anatase TiO2, magnetite Fe3O4, hematite Fe2O3 and monazite (Ce, REE)PO4.
Heavy mineral sands are a class of ore deposits that are an important source of titanium, zirconium, thorium and rare earth elements. This application note describes the use of XRD, together with the Rietveld method and cluster analysis, for phase identification and quantification of heavy mineral sands.
The CubiX3 Minerals industrial diffractometer equipped with a 64-position sample changer, fulfills all modern international safety standards. It can be connected to transport belts for fully automated sample preparation and analysis, including the cleaning and recycling of the sample holders.
XRD can be used to readily identify the mineral phases present in heavy mineral sands, such as ilmenite FeTiO3, rutile TiO2, zircon ZrSiO4, quartz SiO2, anatase TiO2, magnetite Fe3O4, hematite Fe2O3 and monazite (Ce, REE)PO4.
High-speed XRD systems like the CubiX3 Minerals equipped with the X’Celerator or PIXcel detector allow the use of standardless methods such as Rietveld refinement for quality and production control. Cluster analysis of XRD data can facilitate multi-dimensional compositional mapping of ore deposits and drill cores, identifying regions of favorable mineral compositions.
The grain size of the heavy mineral sands is far too large for direct analysis. Samples were therefore ground in a disk mill with a tungsten carbide (WC) vessel and were pressed into steel ring sample holders. The steel rings are compatible with semi-automatic sample preparation equipment with a piston diameter of 35 mm.
The measurements were made using a PANalytical CubiX3 Minerals diffractometer, equipped with the X’Celerator detector and a cobalt tube with incident iron filter. This type of radiation is especially suited to iron- containing materials, as it produces high- resolution data without interference from the sample fluorescence. A measurement time of 5 minutes was chosen to obtain sufficient peak intensities. For good peak to background ratios, a beta filter (Fe) was placed into the incident beam path.
Figure 1. The CubiX3 Minerals diffractometer is equipped with computer controlled slit optics, a variable speed spinner stage and the X’Celerator detector, which produces high- resolution data up to 100 times faster than with traditional instrumentation.
Figure 2. Schematic diagram of the diffractometer configuration
Figure 3. Comparison using the beta filter in the incident or diffracted beam
Clustering or cluster analysis - was conducted on the XRD data collected from the heavy mineral sands. Modern XRD equipment allows rapid collection of hundreds of scans. For exploration and process control applications it would be impossible to analyze each individual measurement. Cluster analysis can be used to greatly simplify data processing by automatically sorting closely related scans into clusters. Representative samples and the two most different scans of each cluster are identified, and outlying patterns are shown. The result is to drastically reduce the amount of data that requires processing.
Two different cluster analyses were performed. In a first test the 64 scans from the reproducibility measurements were analysed. The principal component analysis PCA score plot (Figure 8a) clearly shows two different clusters and one outlier.
The scans of the clusters correspond, as expected, with 26 scans made for testing the reproducibility of the preparation method (blue), and the 37 scans performed on one sample (orange). The different clusters show different distributions in the PCA plot. While the measurements for the precision test are concentrated in a small area, the scans for the different preparations are widely spread in the PCA plot, indicating a higher standard deviation. This demonstrates that within a production control environment, stable or unstable processes or different raw material qualities can be easily recognized.
Cluster analysis can be performed prior to subsequent investigations, such as phase identification and quantification. The most representative scans, and scans that differ the most, can be used as starting points for more detailed examination.
For the purposes of this study, a second test was carried out on several scans from samples taken from different steps during processing (screening). Four different groups and one outlier were calculated (Figure 8b). The clusters corresponded with a higher and a lower concentrated sample and different particle sizes of the sample material.
Figure 8 a,b. PCA score plots for the two different experiments (reproducibility measurements, samples from the different process steps)
Phase identification
Figure 4 shows the phase identification results for one of the heavy mineral concentrate samples. Ten phases were identified including ilmenite, rutile, zircon and hematite, along with smaller amounts of anatase, quartz, actinolite, epidote, diopside and almandine. Phase identification was conducted using PANalytical’s HighScorePlus software. To identify the phases present the software employs a powerful search-match algorithm and compares measured data to reference database PDF4+ of the ICDD.
Batch programs within the software are available to enable automation of phase identification for industrial environments, delivering results at the push of a button.
Phase quantification
Rietveld analysis
Quantitative analysis is possible by various classical methods such as straight line or polynomial calibration with standards. Reliable results can be obtained using the full pattern refinement introduced by H. Rietveld in 1969. The Rietveld method is a modern, highly effective quantification technique. It delivers impressive precision, allows a rapid analysis speed and, importantly, does not require standards or monitors. In addition, the Rietveld method takes into account preferred orientation, lattice parameters, possible solid solutions, crystallite sizes and amorphous contents. The RProfile provides an indication of the mathematical quality of the fit.
Figure 5 shows the results of a Rietveld refinement of a heavy minerals concentrate. The zoomed area - Figure 6 - highlights the fit between measured and calculated profile.
Figure 4. Measurement and phase identification for a heavy mineral concentrate sample (Z = zircon, I = iIlmenite, R = rutile, Q = quartz, Ac = actinolite, Al = almandine, Ana = anatase, Di = diopsite, He = hematite, E = epidote)
Figure 5. Rietveld refinement for a heavy mineral concentrate sample (red dots = measurement, blue = calculation, below = difference plot), RProfile = 3.6
Figure 6. An area of the Rietveld simulation overlaid with measured data
To evaluate the reproducibility of the sample preparation method, a sample was prepared and measured 27 times. Analysis precision was also tested by repeating the measurement of one sample 37 times. The results are illustrated in Table 1. Figure 7 shows the width of distribution (standard deviation or sigma). Two thirds of the results in the plots are within ± sigma, 95% are within ± two sigma and all results are within three sigma.
Table 1. Quantitative results using the Rietveld method for the precision and reproducibility
Figure 7. Standard deviations for rutile and zircon for the precision of the method (levels of confidence: solid = mean, dashes = 1 sigma, dotted = 2 sigma, dashes-dotted = 3 sigma)
CubiX3 Minerals is the ideal choice for fully automated XRD. It features a quick loading mechanism that completes a load/unload cycle in less than 10 seconds - contributing to the overall throughput speed without compromising the accuracy of the analytical result.
Front sample loading simplifies robot arm handling in automated installations and sample transport to and from other equipment can be carried out using conveyor belts, supporting the full integration into automated laboratories.
The CubiX3 Minerals diffractometer is accompanied by software, which is optimized for industrial processes. It allows the implementation of user-defined analytical models and optimization of calibrations based on user standards. It can also be controlled by external automation systems, including the Laboratory Information Management System (LIMS).
Benefits
The Minerals editions of Aeris is PANalytical’s X-ray diffractometer for everyone in the mining industry. Ease of use and maximum benefit are the key aspects. Experience for yourself how the operation of the Minerals edition of Aeris is just a breeze with its intuitive user interface on the built-in touch screen where all results you need are directly displayed.
At the same time, the Minerals edition of Aeris is designed for low cost of ownership. With its low power consumption, virtually unlimited lifetime of the tube and limited infrastructural requirements it guarantees low running costs.
Additionally, Aeris is the first benchtop X-ray diffraction (XRD) system that is fully automatable and can easily be incorporated in industrial production control.
Modern XRD can provide valuable information for mining and process control in the heavy minerals industry through standardless quantification and fast, statistical evaluation of large datasets through cluster analysis. Today’s optics, detectors, and software can provide rapid and accurate analyses suitable for process control environments as well as research.
The Rietveld method has many attributes that lend themselves to process control applications. It is suitable for use with homogeneous and heterogeneous samples, and works with powdered materials. It is relatively fast, cost-effective and able to distinguish between phases that may only differ subtly from one another. Furthermore, the Rietveld method is capable of producing quantitative modal abundances for the phases being analyzed (down to less than 0.5 %) and deals well with overlapping peaks.
Detailed knowledge about the atomic structures of the phases and accurate phase identification are mandatory to achieve optimal results. Modal data can be difficult to extract from samples containing two or more minerals of the same family (e.g. amphiboles). However, provided that the user has a good understanding of the overall approach and philosophy of the Rietveld method, then excellent results can be obtained.
The use of cluster analysis to evaluate XRD data allows fast and reliable tracking of the process. It is highly cost-effective as data evaluation is automatic and does not require dedicated personnel. It is ideal for supporting process and grade control applications as well as exploration.
Sample handling and preparation were critically important for reliable phase quantification. It is recommended that a standardized or automated sample preparation procedure is used.
The data shown here demonstrate that the CubiX3 Minerals system, together with the appropriate software modules, is ideal for the quantitative analysis of heavy mineral sands by XRD. The fast X’Celerator detector was highly valuable for delivering very short measurement times.
