Minerals Engineering 116 (2018) 152–162

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Grade and product quality control by microCT scanning of the world class T
Namakwa Sands Ti-Zr placer deposit West Coast, South Africa: An
orientation study
⁎
A. Rozendaala, , S.G. Le Rouxb, A. du Plessisb, C. Philandera,1
a
Department of Earth Sciences, University of Stellenbosch, South Africa
b
CAF X-ray Facility, University of Stellenbosch, South Africa

A R T I C L E I N F O A B S T R A C T

Keywords: The Namakwa Sands operation is a world class producer of zircon, rutile and ilmenite from mainly semi-con-
Namakwa Sands solidated marine and dune sands. This orientation study shows that the high density contrast between the
Ti-Zr heavy minerals economic minerals as well as the diverse gangue mineralogy allows the use of microCT scanning as a qualitative
MicroCT scanning and quantitative analytical tool. The method has demonstrated the ability to quantify final product quality, grain
size distribution, grain shape definition and the identification of external and internal mineral textures. Grain
size distribution data generated by microCT scans compares well with data produced by other analytical
methods. As a result, the method has the potential to be applied to in situ resource calculations (total heavy
minerals - THM, valuable heavy minerals - VHM), run-of-mine (ROM) grade control and the various stages of the
mineral beneficiation process. MicroCT scanning applied to heavy mineral producing operations appears pro-
mising and should be further assessed against traditional analytical methods. The method also has the ability to
assist sedimentological studies with respect to in situ quantitative grain size and sphericity determinations.

1. Introduction the Late Pleistocene siliciclastic, arenaceous sediments and hosted by

two adjacent ore bodies referred to as Graauwduinen West (GD West)
The Namakwa Sands heavy mineral placer deposit is located near and Graauwduinen East (GD East; Fig. 2). GD West consists of three
Brand-se-baai some 300 km north of Cape Town, along the West Coast strandline-dune couplets set in a transitional shallow marine-aeolian
of South Africa. Tronox Limited controls the operation, which has de- environment, whereas GD East comprises a largely unrelated dune
veloped into a significant, global producer of titanium slag, pig iron, deposit. Superimposed onto the ore-bearing sequence is a pseudo-stra-
ceramics quality zircon and rutile concentrates (Fig. 1). The deposit is tigraphy of duricrust that effectively cemented the bulk of the miner-
classified as a mega-resource with 1.17 billion tons of in-situ ore, con- alized sands to various degrees of hardness. The two ore bodies contrast
taining an estimated 92.6 Mt of total heavy minerals and the equivalent in their bulk chemistry and mineralogy. GD West features noticeable
of 23.9 Mt TiO2 units and 9.4 Mt zircon (Jones, 2009; Philander and better total heavy mineral grades, accompanied by conspicuously high
Rozendaal, 2015a). Currently, mechanized dry mining of mostly semi- proportions of the gangue minerals garnet and pyroxene.
consolidated sand takes place in several open pits, allowing an annual The valuable mineral fraction (VHM) is predominated by ilmenite
production capacity of 23 Mt run-of-mine. Mineral processing capacity (FeTiO3), followed in abundance by leucoxene (an alteration product of
has the ability to deliver 450 kt ilmenite as well as 110 kt zircon and ilmenite), zircon (ZrSiO4) and rutile (TiO2). Their geochemistry, mi-
25 kt rutile annually. The ilmenite is supplied to their titanium smelter neralogy and petrography are variable, but the bulk of the current va-
at Saldanha Bay, producing at least 180 kt of titanium slag and 120 kt luable mineral fraction is recoverable to premium-grade products. The
of pig iron per year (Philander and Rozendaal, 2015a). proximal, Mesoproterozoic Namaqualand Metamorphic Province is
considered the main contributor to the current heavy mineral popula-
2. Geological background tion via established fluvial-marine courses and a fluvial-aeolian cor-
ridor (Philander and Rozendaal, 2015a, 2015b; Rozendaal et al., 2017).
The ore mineral suite is associated with Early Pliocene (∼5 Ma) to

⁎
Corresponding author.
E-mail address: [email protected] (A. Rozendaal).
1
Present address: Tronox Namakwa Sands, South Africa.

http://dx.doi.org/10.1016/j.mineng.2017.09.001
Received 6 June 2017; Received in revised form 22 August 2017; Accepted 14 September 2017
Available online 29 September 2017
0892-6875/ © 2017 Elsevier Ltd. All rights reserved.
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

attritioned to remove siliceous surface coatings.

– Final mineral beneficiation is achieved at the MSP (mineral se-
paration plant; Fig. 3). The two different SCP concentrates are
processed in two different circuits, namely the Magnetic Circuit and
Non-magnetic Circuit to produce the ilmenite, zircon and rutile
products. The Magnetic Circuit employs drum magnetic separators,
followed by High Tension Roll (HTR) electrostatic separators to
reject garnets and other silicates to produce the ilmenite product.
The Non-magnetic Circuit consists of four sub-circuits. The non-
magnetic concentrate is upgraded by Induced Roll Magnetic Se-
parators (IRMS) by rejecting the moderately to strongly magnetic
minerals. Hepworth Acid Leaching (HAL) is used to remove mineral
surface coatings, which suppress electrostatic separation and is fol-
lowed by mechanical attritioning. In a following stage, the Wet
Gravity Circuit removes low-density gangue to produce a high-grade
primary concentrate, and a secondary lower grade concentrate.
Final separation takes place in the Dry Mill Circuit that houses a
combination of multiple stages of electrostatic and magnetic se-
parators to produce a prime grade zircon and lower quality zircon
product (Zirkwa™) and a prime rutile and lower quality Tiokwa™
product (Fig. 4). The final ilmenite, zircon and rutile products are
railed to Saldanha Bay, where the ilmenite is smelted, whilst the
zircon and rutile are stored prior to export shipping (Fig. 1).

4. Aims and objectives

To ensure the reliability of in situ resource estimates, grade of run-

of-mine plant feed, effectiveness and efficiency of the various stages
during the mineral separation process and quality control of the final
products, the mine employs a diversity of analytical methods. THM
content is determined by heavy liquid separation. Optical microscopy-
assisted mineral point counting is used to identify and quantify mineral
Fig. 1. Locality of the various operations associated with the mining and beneficiation of
phases during various stages of the resource estimate, mining and mi-
the Namakwa Sands heavy mineral deposit along the West Coast of South Africa.
neral separation. Sieve analyses (mechanical screening) to determine
grain size distribution is commonly used. XRF quantifies titanium and
3. Minerals processing zirconium content as well as deleterious elements such as silica, iron,
phosphorous, uranium and thorium from primary ore to final product.
The production of saleable minerals from a heavy mineral deposit XRD is used for mineral phase identification and quantification.
requires the physical separation of individual mineral grains. Scanning electron microscope (SEM) based methods such as Mineral
Throughout the text, grains refer to individual sand grains on mm scale. Liberation Analyser (MLA) and QEMSCAN assist with mineral phase
Separation is accomplished by exploiting differences in four mineral identification, mineral textures, gangue-mineral intergrowth relation-
characteristics: density, grain size, magnetic susceptibility and elec- ships and their quantification as well as grain size distribution. The SEM
trical conductivity. The complexity of the ore characteristics observed is also used to study surface coatings of ore minerals. EMPA and LA-ICP-
at Namakwa Sands, translates into a challenging and variable proces- MS are methods used to identify both major and trace element chem-
sing response. Only mineral grade, liberation, magnetic deportment, istry of mineral phases and their inclusions or intergrowths.
grain size and grain chemistry are considered as meaningful mineral Although some of the instrumentation for the more routine analy-
recovery drivers (Philander and Rozendaal, 2014). The successful mi- tical methods are physically based on the mine and may have an ap-
neral beneficiation process at Namakwa Sands is performed at four proximate turnaround time of 24 h to deliver data, the more sophisti-
different plants as demonstrated by the flow chart in Fig. 3. cated off-site methods required significantly longer periods to generate
results and effect implementation. The aim of this orientation study is
– Primary concentration (PCP) by wet spiral separation of the −1 mm to introduce and test a relatively new method, microCT (X-ray micro
+45 µm screened fractions takes place at two plants proximate to Computed Tomography) which allows differentiation between objects
the mine face. The total heavy mineral concentration (THM) of the and minerals based on their density and in some cases trace element
run-of-mine feed is increased tenfold by this process. Duricrust ce- contrast (Ketcham and Carlson, 2001; Cnudde and Boone, 2013; Parian
mented oversize feed is reduced by autogenous milling. et al., 2015). Its use in minerals processing especially for granular mi-
– Secondary concentration (SCP) occurs at a centralised plant at the nerals has been demonstrated initially by Miller and Lin (2004) and
mine site. Low Intensity Magnetic Separators (LIMS) remove highly particle shape analysis demonstrated by Lin and Miller (2005) and Zhao
magnetic material (essentially magnetite) from a PCP East and PCP and Wang (2015). It is considered in this case study that the method has
West concentrate blend. In a next phase, Wet High Intensity the potential to provide valuable high resolution quantitative data in
Magnetic Separators (WHIMS) separate the resultant non-magnetic 3D relating to the heavy minerals industry. It can contribute towards
material in a magnetic fraction, containing mainly ilmenite and phase identification, grain size and shape distribution as well as in-
garnet and a non-magnetic fraction comprising zircon, rutile, leu- ternal and external textural features of ore and gangue minerals. The
coxene and low-density gangue. Further WHIMS treatment of the data can be generated off site or potentially on site by a one-step, non-
magnetic fraction produces a Magnetic Concentrate, whereas a wet destructive process, which should be able to reduce turn-around time,
gravity section upgrades the non-magnetic fraction to produce a improve overall productivity and production cost. This orientation
Non-magnetic Concentrate. Both the SCP concentrates are study tested four samples from the mine, three of which are final

153
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

Fig. 2. Distribution and total heavy mineral grades (THM)

of the extensive upper part (Red Aeolian Sands, RAS) of
Namakwa Sands placer deposit along the West Coast of
South Africa. (For interpretation of the references to colour
in this figure legend, the reader is referred to the web ver-
sion of this article.)

Fig. 3. Simplified process flow sheet used at Namakwa

Sands mineral processing operation. (Philander and
Rozendaal, 2014, 2016).

154
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

Fig. 4. Run-of-mine ore from the Namakwa Sands mine (A).

Total heavy mineral content is approximately 10%. Ilmenite
concentrate produced as feed stock for the titanium smelters
in Saldanha Bay (B). Prime rutile concentrate (C) and prime
zircon for the export market (D). (Optical microscopy, in-
cident light, images taken at the same magnification).

Fig. 5. (A) Prime zircon concentrates showing high quality and purity. Small dark spots Fig. 6. Prime rutile concentrates with bright, high density grains of zircon (A). Contrast is
are low density voids/regions consisting mainly of apatite and diverse fluid or gas. enhanced to visualize rutile. The zircon distribution is accentuated by the 3D image.
Contrast is enhanced to visualize zircon. (B) Three dimensional image showing dis- Mineral identification is confirmed by the elongated, tetragonal shape of the grains and
tribution of fine, high density grains, confirmed as monazite, in the concentrate. MicroCT SEM analyses (B). MicroCT voxel size 1.5 µm, field of view 3 mm.
voxel size 1.5 µm, field of view 3 mm.

155
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

diverse consisting of the valuable heavy minerals (VHM) which include

ilmenite, leucoxene, rutile and zircon as well as heavy minerals with no
or little economic value. These fractions combined define the THM
suite. Minerals with density of less than 2.9 g/cm3 such as quartz and
feldspar may also be present. The selected samples have been subjected
to chemical (Namakwa Sands mine laboratories) and QEMSCAN ana-
lyses (Exxaro Laboratories, Pretoria) for comparative purposes with the
MicroCT results. Sieve analyses also referred to as mechanical screening
was done at Namakwa Sands mine laboratories following international
protocols and controls.
Sample size was reduced by means of a riffler and micro splitters in
order to obtain a representative sample for scanning. MicroCT scanning
was performed at the Stellenbosch University CT facility which has
been described in detail by du Plessis et al. (2016). In order to obtain
best quality images at highest possible resolution, settings were selected
to optimize quality rather than processing time, using methodology
described in du Plessis et al. (2017). Shorter scanning and analysis
times could be possible once the methodology has been streamlined for
this particular application. Each sample was placed in a small tubular
transparent container with dimensions 3 mm diameter and 7 mm depth.
Once appropriate settings were selected ensuring good X-ray para-
meters, each sample was then scanned for a period of 120 min within a
General Electric Phoenix Nanotom S scanner at 120 kV with a current of
80 uA at a resolution of 0.0015 mm (1.5 µm). The X-ray spot size was
limited using the “mode 1″ option, ensuring the X-ray spot is kept below
the voxel size selected. Stepwise rotation of the sample was performed
in 3600 steps in a full rotation of the sample. At each step, the first
image acquired was discarded to ensure no movement artefacts are
present and the subsequent 3 images averaged to produce higher
quality acquisition. Each individual image acquisition time was 500 ms.
Detector shift was activated which allows reduced central rotation axis
artefacts by moving the detector horizontally between step positions.
Reconstruction was performed with system-supplied Datos reconstruc-
tion software using a moderate beam hardening correction factor.
Post processing of the data and advanced image analysis was per-
formed with Volume Graphics VG Studio Max 3.0 software where the
advanced surface determination function allowed the identification and
quantification of high density grains and their volumes. In the advanced
surface determination, the edge of each grain is accurately determined
based on local optimization of the gradient between background and
material in the 3D image data. The result allows not only images of the
grains and their internal details, but also quantitative analyses such as
grain size distribution, largest grain size, number of grains, number of
inclusions and shape analysis.
The detailed methodology to calculate mineral phase concentrations
and grain size distribution included the use of the volume analyser tool
which makes it possible to calculate the volume of the original grains
and high or low-density grains relative to the bulk grain volume. These
volumes can be expressed as weight per cent by using the average
density of each mineral’s phase. In addition to the above relative con-
Fig. 7. Ilmenite concentrate consist of a diversity of grains with variable composition as centrations, which has been applied and described for another appli-
indicated by the contrasting grey scale (A). Contrast is enhanced to visualize ilmenite. At cation by le Roux et al. (2015), grain size analysis was performed using
high magnification the complex internal textures of ilmenite as well as low density in- an advanced methodology as follows. The newly implemented “foam
clusions are clearly visible (B). The concentrate also has scattered grains of zircon based structure analysis” module of VG Studio Max 3.0, is meant for analysing
on the elongate grain shape and SEM analyses (C). MicroCT voxel size 1.5 µm, field of
foam structures i.e. large air voids connected to one another. In this
view 3 mm. Magnified image field of view 1 mm.
case the analysis was run on material instead of background as the si-
tuation relates to sand grains in contact with each other. In this case
products and one a PCP concentrate from the RAS part of the GD East each grain is identified as an individual entity, even though they are
ore body to demonstrate “proof of concept”. touching each other in the container. A simpler methodology which has
been demonstrated by Rothleitner et al. (2016) and Zhao and Wang
5. Methodology (2015) makes use of grain mounting in resin or glue to ensure they do
not touch each other during microCT scanning, which makes the ana-
Three of the four samples studied were randomly collected from the lysis simpler but with a much reduced number of grains. In our work
prime zircon product, prime rutile product and ilmenite product stock grains are placed in a container “as-is” and digitally separated. The
piles and one sample from the PCP East heavy mineral concentrate advantages are that a much larger number of grains are analysed in one
which serves as feedstock to the SCP (Fig. 3). Feed stock mineralogy is scan, less bias is present in grain selection and simplicity of sample

156
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

Fig. 8. Synoptic ternary diagram showing chemistry of titanium phases associated with the Namakwa Sands deposit. Ilmenite and rutile/anatase products are not homogeneous and
display significant variation in chemical composition (Rozendaal et al., 2017).

Fig. 9. CT images of the SCP feedstock which allowed the discrimination between the three dominant phases of the VHM fraction as well as accessory monazite. Grain size variation
between the phases is clearly visible. Monazite – pink, zircon – green, ilmenite – blue and rutile – red. MicroCT voxel size 1.5 µm, field of view 3 mm. (For interpretation of the references
to colour in this figure legend, the reader is referred to the web version of this article.)

preparation. The disadvantage could be potential errors in digital se- The final mineral product samples were used to standardize/cali-
paration of grains, which could be assessed visually to ensure accuracy. brate the grey scale images against the mineral phase such as rutile,
This is also a time consuming analysis, requiring significant computing zircon and ilmenite. Grains of lower density were classified as gangue
power. In both methods the grain volume, surface area, sphericity and and those of higher density than zircon as minerals of potentially high
other parameters are calculated for every individual grain and reported radioactivity such as monazite.
in a spreadsheet.

157
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

analyses. Low density gangue is effectively absent indicating good se-

paration. The concentrate has 99.5% zircon content as calculated vo-
lumetrically and is confirmed by chemical and Qemscan analyses which
indicate a 99.9% purity by weight. Previous work by Philander and
Rozendaal (2014) on concentrates supports this result.
The prime rutile concentrate shows a fair abundance of high density
grains confirmed as mainly zircon by SEM analyses (Fig. 6). Rutile
grains are heterogeneous in composition as seen in Figs. 4C and 6 and
this is confirmed by the subtle variation of grey scale in the microCT
scans. Some of the grains are confirmed as leucoxene. Accessory grains
of higher density confirmed as monazite, also occurs in the concentrate.
The separation of rutile and zircon appears to be less effective and is
demonstrated by the 97.2% by volume rutile content of the prime
product. When analyzed by the QEMSCAN method the rutile con-
centrate has a 98.9% purity by weight when leucoxene is included in
the calculation. This result suggests that some phases such as low ti-
tanium-high silica leucoxene were not included with the rutile popu-
lation in the microCT scan.
The ilmenite concentrate contains minor scattered zircon grains
(Fig. 7). The ilmenite grains are however heterogeneous as shown by
variable shades of grey. Single grains show complex internal structures
and host several inclusions many of which are apatite (Fig. 7B). This is a
typical characteristic of ilmenite as demonstrated by Philander and
Rozendaal (2015a) and Rozendaal et al. (2017) (Fig. 8). As a con-
centrate it has a volumetrically measured 99.9% ilmenite content. This
high purity grade has been confirmed by chemical and Qemscan ana-
lyses and previous work (Philander and Rozendaal, 2014). Grain size is
variable compared to the uniform size distribution of the zircon and
rutile concentrates and is a function of variable chemistry and conse-
quently density.
The SCP feedstock from the GD East part of the mine is a heavy
mineral concentrate consisting of the VHM fraction and minor non-
economic minerals. The microCT scan was able to distinguish the three
dominant phases zircon, rutile and ilmenite (Fig. 9). Concentration by
volume of the zircon (28.6%), rutile (24.2%) and ilmenite (47.0%)
indicates that the feedstock consists almost entirely of the VHM frac-
tion. Trace amounts of monazite are present (< 0.2%). Grain size is
variable and reflects what has been observed in the individual con-
centrates (Cornelius et al., 2014). The VHM concentration results are
poorly supported by QEMSCAN analyses (zircon 21%; rutile 5%; il-
menite 73%) and are not considered reliable. This indicates that
microCT analysis clearly require refinement when samples with mul-
tiple mineral phases are considered. The discrimination between mi-
nerals with variable chemistry and alteration products such as ilmenite
and rutile for example requires attention.

6.2. Grain size

The CT images have visually shown the grain size variation between
samples and also within the SCP feedstock. The results of these scans
allowed the computation of grain size distributions. The method makes
use of an inverted foam structure analysis, digitally separating grains
and providing for each grain volume, surface area, sphericity and other
quantitative values. In this work we could calculate an effective dia-
Fig. 10. (A) Prime rutile concentrate. (B) Colour coded volume analysis of rutile con-
meter of each grain based on the volume and calculating the diameter
centrate. (C) 3D image of colour coded volume analysis. MicroCT voxel size 1.5 µm, field
of a sphere with the same volume, referred to as the equivalent sphe-
of view 3 mm. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of this article.) rical diameter. An example is shown for rutile, in Fig. 10A where the CT
slice image shows the grains, whilst Fig. 10B shows the colour coded
analysis results for volume of each grain. The results are then presented
6. Results
as frequency and cumulative frequency diagrams and compared to
grain size measurements as performed by the Namakwa Sands mine of
6.1. Concentration
the same fraction by means of grain size screening and QEMSCAN. The
comparative results are graphically presented in Fig. 11 and statistical
The microCT scan of the prime zircon concentrate shows an image
data in Table 1. The excellent correlation between the CT results and
(Fig. 5) consisting almost entirely of homogenous zircon grains with an
the other two methods for each mineral phase as indicated by the d50
occasional scattered high density grain, confirmed as monazite by SEM
for example, is clear and indicates that quantification of grain size by

158
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

Fig. 11. Histograms showing distribution of grain size for zircon, ilmenite and rutile as determined by mechanical screening, QEMSCAN and Computed Tomography. Of the three
methods, results from the microCT analyses proved to be the most realistic from a depositional environment perspective of the placer deposit.

means of microCT scanning is a feasible alternative. Grain size dis- than the QEMSCAN and mechanical screening methods as reflected by
tribution for the various mineral concentrates shows that on average the steep slope of the cumulative frequency curve (Fig. 11). From a
rutile is slightly coarser grained than zircon and ilmenite. This char- sedimentological perspective a narrow grain size distribution is an in-
acteristic relates to its lower density and behaviour under conditions of dication of high maturity of detritus normally associated with reworked
hydraulic equivalence in fluvial and aeolian depositional environments: aeolian and to a lesser extent marine sediments (Pettijohn et al., 1987;
small dense grains behave similarly to large less dense grains under Force, 1991). The Namakwa Sands deposit has this particular associa-
similar depositional conditions (Force, 1991). The narrow range and tion and it concluded that the microCT scan results are more realistic
good correlation of grain size of the various mineral products is a result than shown by the other methods (Philander and Rozendaal, 2015a,
of screening the SCP feedstock between +45 and −180 µm prior to 2015b). This can be attributed to the fact that grain size distribution by
entering the MSP. Grain size is an important parameter and is a defi- microCT scanning considers 3D images and consequently volume in its
nitive characteristic of the ore body. It plays a critical role in the ben- calculations thus eliminating stereological bias as imposed by 2D
eficiation process: for example a narrow range is conducive to better QEMSCAN methodology.
recovery. Grain size distribution of individual phases in the SCP feed was
Although results of the microCT analyses compare well to the other difficult to determine and the present method requires some develop-
methods its grain size distribution of the three fractions is narrower ment to allow this to be done with confidence.

159
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

Table 1 Table 2
Results of comparative grain size study between mechanical screening, QEMSCAN and Statistical parameters of sphericity histograms of the three final products.
Computed Tomography. Grain size in µm.
Statistical Parameter Zircon Ilmenite Rutile
Statistical Parameter Zircon Ilmenite Rutile
Mean 0.56 0.54 0.54
Mechanical Screening Mean 105 108 109 Median 0.56 0.55 0.55
Standard Error 0.28 0.33 0.26 Standard Deviation 0.04 0.04 0.04
Median 98 98 116 Sample Variance 0.00 0.00 0.00
Standard Deviation 21 25 19 Kurtosis 2.60 6.68 6.08
Sample Variance 441 618 377 Skewness −0.42 −1.49 −1.36
Kurtosis 1.89 6.17 1.35 Range 0.43 0.68 0.66
Skewness 0.61 1.75 0.13 Minimum 0.38 0.13 0.15
Range 209 209 209 Maximum 0.81 0.81 0.81
Minimum 23 23 23 Count 11,597 12,275 10,678
Maximum 231 231 231

QEMSCAN Mean 104 106 104

Standard Error 0.30 0.38 0.31
Median 105 105 105
Standard Deviation 30 38 31
Sample Variance 922 1 453 931
Kurtosis 1.74 4.08 3.49
Skewness 0.48 1.24 0.53
Range 260 280 280
Minimum 15 15 15
Maximum 275 295 295

Computed Tomography Mean 108 103 110

Standard Error 0.18 0.19 0.19
Median 105 100 110
Standard Deviation 20 21 19
Sample Variance 386 452 379
Kurtosis 3.83 4.92 4.36
Skewness 0.61 0.97 −0.15
Range 237 269 306
Minimum 6 6 6
Maximum 243 275 312
Fig. 13. Low density inclusions in ilmenite (dark grey) which could be silicate minerals
reducing the quality of the final product. MicroCT voxel size 1.5 µm, field of view 1 mm.
6.3. Grain shape
(Philander and Rozendaal, 2015b). From a geometallurgical perspec-
Variation in grain shape of the four samples is clearly visible in the
tive grain shape has an influence on recovery. It has been observed that
images provided by the microCT scans (Figs. 5–7 and 10). Grain shape
elongated grains would be more susceptible to be rejected on roll se-
expressed as a function of sphericity was calculated for each mineral
parators, on spirals and even screens. As a result sphericity indices re-
fraction by using measured surface area and volume of individual
veal characteristics of the ore body and can be predictive in terms of
grains measured from microCT data using the described methodology.
expected recovery.
The results are presented in a series of frequency diagrams and de-
This type of data also allows its use in other disciplines such as
monstrate that in general the mineral concentrates have a similar
sedimentology.
sphericity, with zircon having a slightly higher index (Fig. 12; Table 2).
MicroCT scans (1.5 µm voxel size) also provided images demon-
The general low sphericity is the result of the primary elongated tet-
strating surface features of single grains, intergrowths and internal
ragonal/trigonal crystal shape of the mineral phases. The narrow range
structures. Low density surface-connected minerals inside zircon grains
is attributed to the sediment maturity and indicates that the mineral
were discernible (Fig. 14A). This type of impurity is a common chal-
phases have been derived from and subjected to the same sedimento-
lenge on the mine. Quantification of the percentage coverage of the
logical environment and depositional conditions. Sphericity range of
grain surface should be possible. Aggregates of zircon or any other
zircon is the lowest because it is the most sedimentologically mature as
valuable heavy mineral, the result of lockings with duricrust should be
indicated by its diversified provenance spanning billions of years

Fig. 12. Histograms showing the sphericity of the three final products.

160
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

methods such as mechanical screening or QEMSCAN of the same

samples. Mineral size fraction that is recovered is set between +45 and
−180 µm and grain size analyses of ROM (run-of-mine) will provide an
indication of potential ore losses imposed by screening. Quantification
of the latter is possible but the methodology will have to be refined in
future work. Grain shape can also be determined by CT scanning and is
expressed as a function of sphericity. Low sphericity elongated minerals
such as zircon and rutile are often lost in the screening (classification)
or other processes, resulting in a loss of VHM recovery. It is possible
that this loss can be quantified.
External and internal grain characteristics are important parameters
in the mineral separation process. Surface connected inclusions, parti-
cularly if mineral composition is of low density, can be identified and
potentially quantified. These inclusions impair the physical character-
istics of a mineral, like zircon for example, which will result in poor or
non-recovery in the MSP circuit. Similarly gangue-ore aggregates can
be identified and potentially quantified. Internal textural features such
as intergrowths, exsolution features and alteration can be identified
with high resolution scans and provide an indication of the complexity
of the individual mineral grains. Abundant inclusions in zircon such as
apatite for example may contribute to unwanted elements in the con-
centrate imposing a penalty at the sales end. Although this cannot be
quantified with the microCT scans as yet, it can indicate the possible
cause and guide potential solutions. At this stage the QEMSCAN and
MLA analytical methods are clearly better suited to identify mineral
phases and their intergrowths at high resolution, although the latter
only in two dimensions. These methods presently supplement microCT
studies and provide the required mineral identification used as standard
reference material.
The orientation study has also highlighted the limitations of
microCT at its current stage of development. Visually, in particular in
3D and qualitatively the results indicate significant potential as an
analytical tool in the heavy minerals industry. Samples with relatively
simple mineralogy work well and provide comparable quantitative re-
sults with other methods such as mechanical screening and QEMSCAN.
Mineralogically and texturally complex samples however present a
Fig. 14. Prime zircon product showing a diversity of low density internal and surface
connected inclusions (A). Zircons as impurities in the prime rutile concentrate showing challenge and the method requires further development to bring it in
diagnostic concentric zoning (B). MicroCT voxel size 1.5 µm, field of view 1 mm. line with current conventional analytical methods such as XRF geo-
chemistry and SEM based methods which include QEMSCAN and MLA.
The Namakwa Sands deposit per se offers this opportunity and further
clearly visible and could potentially be quantified but has not been
research will ensure an expanded and reliable microCT scan database
attempted in this study. The ability to demonstrate the internal struc-
for such and similar deposits. This will be required to demonstrate and
ture of complex minerals such as ilmenite can be seen in Fig. 13. Si-
develop the broad based application of the microCT scanner in the
milarly are minerals such as apatite, fluid or gas inclusions in zircon or
minerals industry.
rutile visible on these scans (Fig. 14). Such data will provide an in-
dication of deleterious impurities that are difficult to remove by mi-
8. Conclusions and recommendations
neral separation. High resolution scans show concentric zonation of
high density minerals which is typical of zircon.
It has been demonstrated that microCT scanning technology has the
potential to make a meaningful contribution to the mining and bene-
7. Discussion ficiation sections of the heavy minerals industry. It has the potential to
quantify several critical parameters based on 3D information and sta-
The orientation study using four samples from the Namakwa Sands tistically meaningful database in a non-destructive one-step process.
mine demonstrated that the microCT scanner has the capability to The orientation study will have to be expanded to include run-of-mine
measure and quantify mineral phases and parameters that are critical in ore from both West and East ore bodies and material from the various
the heavy mineral industry. Individual mineral phases can be identified stages in the mineral separation process to ensure representivity
by using standards and concentration as volume either as gangue or ore (Fig. 3). The results will have to be compared to those of the presently
types quantified. By converting volume to weight percent will allow the used conventional methods such as XRF, QEMSCAN, MLA and precision
measurement of ore grade either as valuable heavy minerals (VHM) or and accuracy demonstrated to justify its position as a supplementary
total heavy minerals (THM), the industry standards. In addition, in- and possibly alternative analytical tool in the heavy minerals industry.
dividual mineral concentrations can also be measured during various The viability of the method with respect to measuring and data pro-
stages of the separation process (Fig. 3). Identification of very high cessing time, a function of the instrument capacity should be de-
density, deleterious radioactive minerals such as monazite is possible termined and compared to the present protocol on the mine. Finally a
and concentration can be quantified. cost-benefit analysis should be conducted to demonstrate the ad-
Scanning results allow calculation of grain size: an important vantages of the new method. Provided the abovementioned outcomes
parameter particularly with respect to the in situ ore and PCP heavy are positive, development of the capacity to do real time online mea-
mineral concentrates. The results compare well with other analytical surements at various stages in the mining and beneficiation process

161
A. Rozendaal et al. Minerals Engineering 116 (2018) 152–162

should be considered. ore using X-ray microCT: case study. Comput. Geosci. 85, 75–80.
Lin, C.L., Miller, J.D., 2005. 3D characterization and analysis of particle shape using X-ray
The method has also demonstrated its potential application in other microtomography (XMT). Powder Technol. 154 (1), 61–69.
areas of geology such as sedimentology where quantitative data of grain Miller, J.D., Lin, C.L., 2004. Three-dimensional analysis of particulates in mineral pro-
size, sphericity and roundness are imperative in lithostratigraphy and cessing systems by cone beam X-ray microtomography. Miner. Metall. Process. 21
(3), 113–124.
basin analyses. Parian, M., Lamberg, P., Möckel, R., Rosenkranz, J., 2015. Analysis of mineral grades for
geometallurgy: combined element-to-mineral conversion and quantitative X-ray dif-
References fraction. Miner. Eng. 82, 25–35.
Pettijohn, F.J., Potter, P.E., Siever, R., 1987. Sand and Sandstones. Springer-Verlag, New
York 553p.
Cornelius, L.J., du Plessis, A., le Roux, S.G., Rozendaal, A., 2014. Grade determination of Philander, C., Rozendaal, A., 2014. A process mineralogy approach to geometallurgical
mineral sands using X-ray CT: proof of concept. Research report, CT Facility model refinement for the Namakwa Sands heavy minerals operations, west coast of
Stellenbosch, Stellenbosch University, p11. South Africa. Miner. Eng. 65, 9–16.
Cnudde, V., Boone, M.N., 2013. High-resolution X-ray computed tomography in geos- Philander, C., Rozendaal, A., 2015a. Geology of the Cenozoic Namakwa Sands heavy
ciences: a review of the current technology and applications. Earth-Sci. Rev. 123, mineral deposit, west coast of South Africa: a world-class resource of titanium and
1–17. zircon. Econ. Geol. 110, 1577–1623.
du Plessis, A., le Roux, S.G., Guelpa, A., 2016. The CT Scanner facility at Stellenbosch Philander, C., Rozendaal, A., 2015b. Detrital zircon geochemistry and U-Pb geochro-
University: an open access X-ray computed tomography laboratory. Nucl. Instrum. nology as an indicator of provenance of the Namakwa Sands heavy mineral deposit,
Methods Phys. Res., Sect. B 384, 42–49. west coast of South Africa. Sed. Geol. 328, 1–16.
du Plessis, A., Broeckhoven, C., Guelpa, A., Le Roux, S.G., 2017. Laboratory X-ray micro- Philander, C., Rozendaal, A., 2016.The geometallurgy of Namakwa Sands: challenges and
computed tomography: a user guideline for biological samples. GigaScience 6 (6), solutions. In: 35th International Geological Congress, Cape Town South Africa.
1–11. Rothleitner, C., Neuschaefer-Rube, U., Illemann, J., 2016. Size and shape determination
Force, E.R., 1991 Geology of titanium-mineral deposits. Spec paper 259, Geol. Soc. Am., of sub-millimeter sized abrasive particles with X-ray computed tomography. In: 6th
Boulder Colorado. Conference on Industrial Computed Tomography (iCT), p. 2016.
Jones, G., 2009. Mineral Sands: An Overview of the Industry. Iluka Resources limited, pp. Rozendaal, A., Philander, C., Heyn, R., 2017. The coastal heavy mineral sand deposits of
1–26. Africa. S. Afr. J. Geol. 1 (120), 135–152.
Ketcham, R.A., Carlson, W.D., 2001. Acquisition, optimization and interpretation of X-ray Zhao, B., Wang, J., 2015. 3D quantitative shape analysis on form, roundness, and com-
computed tomographic imagery: applications to the geosciences. Comput. Geosci. 27 pactness with μCT. Powder Technol. 291, 262–275.
(4), 381–400.
Le Roux, S.G., du Plessis, A., Rozendaal, A., 2015. The quantitative analysis of tungsten

162