Geometallurgy and automated mineralogy -

A tool for ore deposit evaluation, prediction of processing

problems, and scoping process improvements ahead of and
during mining

T. Wallmach, M. Rousseau, and S. Blancher

ERAMET Ideas, France

Increasing competition in the minerais industry and fluctuating coramodity prices require
new ways of saving energy, lime, and general operational costs. A good understanding of
physical processing or pre-processing streams that can potentially cut these costs requires
detailed analyses of chemical and physical behaviours and processing responses during
rainera]. processing. It is very useful to perform a detailed mineralogical and micro-textural
characterization of materials (ore, tailings, and waste) that addresses, among other
parameters, particle and grain sizes, as well as particle densifies. The choice and/or
corabination of the 'best' processing approaches is crucial for processing efficiencies, and
can be established and verified by using automated mineralogy with the associated
software. A sample of low-grade iron ore from El Volcan, Mexico, serves as an example to
demonstrate in a step-by-step approach how QEMSCAN® analyses provide processing
information. Elements under consideration include iron, phosphorus, and sulphur.

INTRODUCTION

Cheraical investigations in the past did not include detailed mineralogical information_ By including
this information it is possible to predict the process behaviour of minerals in ore deposits. This can raake
the difference between an ore resource of potential economic interest and ore reserve of definite
economic value. Mine planning is also increasingly undertaken using mineralogical and process-related
factors that can be simulated by means of the QEMSCAN® software. A further field of application is the
recycling potential of tailings and waste materials to recover minerais and specific elements of value.
The literature provides numerous examples of applications and case studies as to how automated
mineralogy can add value to process improvement. In the past few years a large number of papers and
communications have been published on the interactions between automated mineralogy, ore
processing, and georaetallurgy. As we do not aim to provide a literature review, only a few
representative publications that address the use of automated mineralogy during ore processing are
mentioned. Examples are Latter et ai. (2011), Schouwstra and Smit (2011), Baum (2014), Gu, Schouwstra,
and Rule (2014), Zhou and Gu (2016), and Becker, Wightman, and (2016).

EXPERIMENTAL AND ANALYTICAL METHODS

Wlineralogical characterization was performed using QEMSCAN®, scanning electron microscopy, and
electron microprobe analyses. QEMSCAN® is an automated technique for the fast characterization of
rainerai and noncrystalline phases on polished sections by means of scanning electron microscopy using
energy-dispersive spectroraetry (SEM-EDS). QEMSCAN® analyses for this project were carried out at
ERAMET Research Wlineralogical Services, using an FEI Quanta 650F SEM platform with two Bruker
Xflash 30 mm silicon drift energy-dispersive X-ray detec tors.

Geometallurgy Conference 2018, Rack to the Future

Cape Town, 6-8 August 2018
The Southern African Institute of Mining and Metallurgy
7
For the QEMSCAN® analyses in this study 77 000 particles were analysed on a micrornetre scale. Each
analysis generating an EDX spectrum took 2.3 milliseconds. A total of 14 million analyses are therefore
available for minerai classifications and microtextural investigations. The total analytical lime
arnounted to 9 hours.

INTRODUCTION TO THE QEMSCANt ANALYTICAL VIETHOD

QEMSCAN® is an automated mineral and phase characterization software package combined with a
scanning electron microscope (SEM). Energy-dispersive spectrometry (EDS) signais are used to
characterize minerais and phases. The characterization addresses physical, chemical, and textural
properties of minerais and phases in order to statistically evaluate this information, individually or in
corabination, on individual grains or particles as well as on the entire sample.

The electron beam of the SEM scans the surface of each prepared polished section on a micrometre-by-
raicrom.etre basis. EDS detectors collect X-rays emitted by the sample. The EDX spectra provide the
essential data for minerai or phase identifications. The data obtained is compiled in a comprehensive
database. QEMSCAN® analyses also collect information on the greyscale level of backscatter electron.
(BSE) images (Figure 1) and the analysis rate, which is linked to the surface state. During QEMSCAN®
measurements a large number of microanalyses, typically several millions, is collected, providing
statistically sig:nificant chemical and (alter data processing) mineralogical information.

In this paper we will show how this information can be used for pre-processing ores, and also how EDX
analyses can be used to identify and quantify trace elements that impact on the ore quality.

The creation of a reliable species identification protocol (SIP, the mineralogical database) is one of the
most critical aspects of precise analyses of any sample. Chemistries and densifies of the individual
phases are entered into the so-called primary list in order to convert area analyses into weight
percentages iii reporting options. Figures 1 and 2 show a comparison of a SEM image and a
QEMSCAN®-generated image displaying varions minerais and textural relationships. Ail of these can
be quantified and used for process modelling. The quality of the database can be tested by comparing
the bulk chemical analyses of the sample with the recalculated chemistry based on the mineralogical
composition. A good correlation is presented in Figure 3.

Figure 1. SEN! backscatter electron (B SE) image. White bar = 1 min. Di_ erent greyscale lev& are a function of
the average atomic weight of the elements fonning the minerais. The lighter shacles represent minerais
contaaning more of the lzghter elements.

8
 Background
Quartz
Goethite
Goethite (P)
He matite
1=1 Pyrite
Py rrhotlte

Pee ud orutl I e
Chrom rte
Cha Eco py rite
Chlorite
El Plagioclase
• K Fe hispa r
Biotite
Muscovite
Aa pante
131 Epidote
11 Amphibole
Pyroxene
Zircon
Dl Gamet
Sphene
Tourmaline
El Ta k
El Calcite
ni Skie rite
MonazIte
0 the rs

Figure 2. QE1VISCAN® image of the saine field as presented in Figure 1, showing dient identified and
classified minerais by means of varions (fuse) colours. Each particle is unique in tenus of size, drape, density,
chernistry, and nuneralogy. nus information is essential for (Jetez-mann& the echveness of density separatzon.

50 Minerai Name
AI Z03
Mi Background 0.00
0,01
Quartz 3,87
45 C 0.37
in Goethite 5,41
Ce 3,12
Goethite (P) 0.74
Ce 0,00 Hematite 53,69
40 CI 0.00 ©Pyrite 1.40
Cr 0,06 Pyrrh otite 0,79
F 0,02 Ilmenite 0.02
35
Fe 45,31 P se udo rutile 0,01
Chrorrete 0.12
ris H 0,16
„,~~30
u) K 0.62 Chalcopyrite 0,14
Le 0,00 ■ Ai Chie rite 3.28
LI 0,00 • C Plagioclase 4.48
25 • I< Feldspar 3,34
Mg 1.87 • Ca El Filante 0.68
rn Mn 0.01
• Cr Muscovite 0,64
•
u~~20 Na 0.36
Aapetlte 0.71
0 Nd 0,00 • Fe Epidote 3.22
D 35.64 ■ K Amphibole 10.88
15 0,11 • Na Pyro Ken e 0,28
1.10 Ta-con 0.04
8.87
P Garnet 2,22
10 Si
• Si
Th 0.00 Sphene 0,30
Ti 0.07 • Ti Tourmaline 0.02
Y 0.00 El Talc 0,11
Calcite 2.83
Zn 0,00
Zr 0,02
D Siderite 0,26
MonazIte 0.00
10 15 20 25 30 35 40 45 50 • Others 0,21
Chemical Assay

Figure 3. Chemical assay reconciliation with Qemscano-recalculated element concentrations, as indicated in the
figure, from minerai compositions. A good correlation is necessary ta artifil771 e good miner al identification. Die
znineralogical composition of the investigated semple (in wt%) is presented on the right-land sida of the figure.

9
ORIGEN OF THE ANALYSED WON ORE SAMPLE

The sample was provided by a client in order to demonstrate the QEMSCAN® application for minerai
characterization and to provide process-related information. The particle size of the sample did not
exceed 600 prit

The sample material originates from the El Volcan iron ore deposit in Mexico. El Volcan is a surface
mine located near the town of Rosario in the state of Sonora, owned and operated by Arcelor Mittal.
After closure of more than one year, operations at El Volcan were resumed in the first quarter of 2017
as a result of a revised mine plan following improved price conditions.

The deposit is geologically associated with a rhyolitic dome and rhyolitic lava flows, and is discordant
with the rhyolite. The ore is described as containing magnetite as the main ore minerai, which is locally
replaced by haematite. Iron-rich epitherrnal veins are abundant. The analysed sample is an example of
this replacement, the main ore minerai being haeruatite. The high phosphorus content, which is ascribed
to the presence of apatite, impacts on product quality. The representativeness of the sample cannot be
verified. Nevertheless, the material is near-perfect for performing a geometallurgical pilot study on one
sample that highlights the potential to improve ore qualifies and process options.

STUDY AILVIS

The provided sample lias a Fe concentration of 45 wt%, well below the desired quality target of 60%.
The phosphorus content amounts to 0.11 wt%, thereby being above the critical threshold of 0.10 wt%.
High-quality steels require phosphorous contents of <0.05 wt%. The sulphur content of 1.10 wt% is also
higher than the 0.5 wt% required for steelmaking. Commonly, leaching and acid leaching processes are
applied to reduce elevated concentrations of these elements. The following section demortstrates that
alternative methods, based on physical pre-concentration of the ore, can be applied to reduce the
concentrations of phosphorus and sulphur. The principal considerations in the following discussion can
also be applied to a great number of other ore types.

INCREASING THE GRADE AND RECOVERY OF WON BY PROCESS MODELLING

The QEMSCAN® software has a great number of process modelling options, only a few of which can be
addressed in the framework of this paper. One of these options is modelling density separation of
particles. Each particle can contain several phases or grains of different densifies, chemical
compositions, and sizes. The sum of all these variables is taken into consideration in the modelling
software. Figure 4 presents modelling results at a density cut-off of 4.0 g/cm3. The table accompanying
Figure 4 shows selected elernent concentrations, minerai proportions in mass per cent, mass flow
information, and average particle sizes in each fraction.

The >4.0 gicm3 fraction, which represents 58.8 mass% of the sample, lias a Fe concentration of 66.21
wt% compared to 45.31 wt% in the bulk sample. Sirnulated theoretical grade vs. recovery curves for the
bulle sample and >4.0 g/cm3 fraction are presented in Figure 5. The phosphorus content in the sanie
fraction arnounts to 0.05 wt% compared to 0.11 wt% in the bulle sample. The sulphur content increases
from 1.10 wt% in the bulk sample to 1.75 wt% in the >4.0 g/cm3 fraction. Apart from the elevated
sulphur content, this fraction is a high-quality iron ore. The potential for removing sulphur is addressed
in the following section.

10
 <4.0
Density separation simulation >4.0

k
Cal rd Med ESD Pairle 3se 177,92 231.43
iMal Mass 41,24 58,78
Tats11.4dume 14,01 11.85
Flembrad.ril4mss(916) Fe 15,54 06,21
0,20 0,05
0,17 1,75
20,34 0.82
Minci trtssrld 0ua1e 9,00 027
Gedree 5,30 5,49
G orlhee {F1 1,15 0,45
Hemitte 7,48 88,13
Pyi rle 0,20 2.25
Pyr exile 0,12 1,27
lb-merle 0,03 0,02
Psaalodetle 0,03 0,06
C1strrsto 0,00 0,20
C1lalcopynlo 0,06 0,10
91H Er de 7,01 088
Plerrdase 10,82 0,18
H F sidspa 7,99 0 07
Bride 2,19 0,13
Mescuri 1,54 0,01
AaPul 1,51 0,18
Eperk 7,39 0.30
Ankh [Kra 24,74 1 15
Pr merle 0.53 0,07
Zuccn 0,07 0,01
arnr9 5.05 0,24
Seer* 0.08 0,05
ictrimirne 0,05 0,00
0,25 0,01
ceo 8,40 0,33
laiera 0,31 0,22
Remota 0,01 0,00
5Nym 0,30 0,15

Figure 4. Density fraction modelling at a cut-off of 4.0 g/cni3 shows chemical and mineralogical variations in the
respective fractions. Mass flow infora nation as well as average particle size in each fraction is also presented. The
high-density fraction arnounts to 58.76 rot% of the bulk sample. The phosphorais content is below the threshold
value, but the sulphur content has increased with respect to the original sanrple as sulphides are enriched in the
dense fraction of the sinople.

70 -

68

66 -

64 -

62 -

60 -

al 58 -
-Lm
m •
CD 56 -

54 -

52 -

50 -

48-

46-

44
50 55 60 65 70 75 80 85 90 95 100
Reco ver y

Figure 5. Theoretical grade-recoveiy curves; orange for the bulk iron ore sarnple and blue fer the >4.0 g/cni3
fraction. The blue curve is valid for 58.76 mass% of the bulk sinople.

11
The grade improvement is facilitated by the good liberation of haematite. Figure 6 shows that more than
half of the haematite particles are liberated by more than 90%. The definition of liberation' casing the
QEMSCAN® terminology is elsewhere (e.g. in MLA analyses) referred to as 'minerai locking'. These
terms should not be confused with the terni. 'free surface', winch is often used as a synonym for
liberation, but has a totally different textural connotation.

Liberation of Hematite

o%
r j 10%
<= 20%
— <= 30%
ffl <= 40%
I <= 50%
<= 60%
<= 70%
J <= 80%
— <= 90%
< 100%
— 100%

0 10 20 30 40 50 60
Mass Hematite

Figure 6. Of the 54 wt% haematite in the bulk simple, 42 zid% is more thart 80% liberated (light green and
brown colours in the figure).

THE POTENTIAL TO DECREASE THE SULPHUR CONTENT BY FLOTATION

The above discussion shows that the iron ore can be upgraded to die required specifications for iron
and phosphoras. Nevertheless, the amount of sulphur rerrtains too high at 1.75 wt%, more than three
tintes the desired value of 0.50 wt%. During ore characterization it is often vert' useful to consider the
element distribution between various size fractions. For the density fraction >4.0 g/ cm3 it becomes
apparent that sulphur is partitioned in the size fraction <340 gin. Figure 7 presents the results of iterative
modelling to provide the best cut-off size fraction for sulphur concentrations . The sulphur
concentration of 1.75 wt% can be reduced to 1.04 wt% by removing die <340 prm size fraction. The >340
}Am size fraction represents only 17.16 wt% of the total sample, or 29.24 wt% of the 58.69 wt% of the iron
ore concentrate (>4.0 g/cm3). The <340 Inn size fraction represents 41.53 % of the total sample, with a
sulphur concentration of 2.01 wt%.

12
 size<340 urn ze>3 40 11m
<540 >340
Calcul ated ESD Panic le Size 177.59 394,73
Total M355 41,53 17,16
Total Volume 344
Fe 66,16
0,06
2,01
Si 0,80
CI uariz 0,24 0,26
G oeil te 5,61 5.41
Goeihita (P) 0,44 0,47
Hematite 86,72 87,10
Pyrite 2,7:2 1,00
Pyrrhonte 1,26 1,25
limonite 0.02 0,01
Pseudor ue te 0,00 0,00
Chromita 0,00 0,68
Chalcopyrite 0,23 0,10
Chlorite 0,71 0,50
Plagioclase 0,16 0,18
K Fel dsper 0,08 0,06
Biotite 0.15 0,10
Muscovite 0,01 0,01
Aspatne 0,10
Epidote 0,26 0,35
Amphibole 1,08 1,10
Pyroxene 0,08 0,05
Zircon 0.01 0,01
Gamet 0,10 0,31
Sphana 0,0‘t 0,116
Tourmal i ne 0,00 0,00
Talc 0,03 0,01
Calcite 0.38 0,26
Siderite 0,25 0,16
Monazite 0,00 0,00
°thora 0,14 0,18

Figure 7. Sulphur concentrations in the size fractions <340 pin and >340 !mi showing a clear partitioning of
sulphur with respect to particle size. About twice as mach sulphur is presented in the smaller size fraction.

The mineralogical analysis shows that sulphur is partitioned mainly between pyrite and pyrrhotite
with pyrite being the main host for sulphur as shown in Figure 8.

Mass S
o o o o

Mi Pyrite
Pyrrhotite
Chalcopyrite

Figure 8. Elemental clepor tment of sulphur, showing that pyrite is the main host for sulphur, followed by
pyrrhotite, and finally chalcopyrite.

Pyrrhotite bas a high magnetic susceptibility, which makes this minerai stand out in the overall
nonmagnetic minerai assemblage of the investigated hematite-rich iron ore. Modelling can be used to
indicate the degree to which the sulphur content can be decreased by removing this minerai it is
assumed that a particle containing more than 20 wt% pyrrhotite can be removed in a magnetic field of
adequate strength. The results are presented in Figure 9. This result is based on iterative modelling to
obtain tire best cut-off value. Even though the results meet the target sulphur concentration of 0.5 wt%,
it remains highly questionable whether a magnetic separation process can remove the 1.22 wt% of
particles that contain >20% pyrrhotite.

13
 Partirfes with > 20% pyrrhotite sze<340 len size<340
Pyrrhotite rich <340 >340
Ce lculated ESO Panic le Size 210.43 177,82 393,82
Total Mass 1,22 40,77 1530
Tel Volent 0.20 8.18 9.9 4
Fe 61,37 66,24 66,54
P 0.20 0,05 0.04
24,14 1,55 9.52
Si 0.04 0,80 0,82
Quartz 0,50 0,23 0,20
Grmeite 2.30 5,66 5,52
Ginethite (P) 1,45 0,42 0,45
Heetaine 29,92 86,87 88,42
Pyrite 3,70 2,71 0,93
Pyrrhertite 67,64 0.06 4106
Il men kit 0,08 0,02 0,52
Peaudoruble 0.00 0.00 0,00
Ch rectite 0.00 0,00 0.70
Chalcopyrite 1.38 0,24 0,00
Chlorite 0.00 0,72 0,51
Plagioclase 0.12 0,10 0.14
K Fe 'dalla r 0,03 0,08 0.06
Bipale 0.00 0,16 0.10
Muscoyhe 0.01 0,01 0.01
Aapati le 0,65 0,16 0,10
Spi:lote 0,07 0,20 0,37
Amphibole 2,00 1,10 1,22
Py roxene 0,01 0,09 0,05
Zircon 0,01 0,01 1101
Gomel 0,08 0,19 0,31
Sphane 0,03 0,04 0.07
Tourmaline 0,00 0,00 len
Talc 0,00 0,03 0,01
Calcite 0.01 0,39 0.27
Siderite 0,01 0,26 0.16
Morte rite 0,00 0,00 0.00
❑trera 0,05 0,14 0.18

Figure 9. Magnetac sepiaration modelling of pyrrhotite removal in combination with particle size fractionation.
Particles with more thon 20 wt% pyrrhotite are presented in the kft-band side of the figure.
The >340 /am size fraction without pyrrhotite particles Fias a sulphur content of 0.52 wt%, which is very close to
the Larget of 0.5 wt%.

THE PRESENCE »al PARTITIONING OF PHOSPHORUS

The role of phosphorus in iron ores merits a section on its own_ In most cases phosphorus is ascribed to
the presence of apatite, occurring as discrete grains or as microcrystallites in goethite. The study
investigated the partitioning of phosphorus by automated mineralogy. QEMSCAN® analyses are
mainly used for minerai identification but are rarely used for direct chemical quantification as the EDX
acquisition times of several milliseconds are considered to be too short for obtaining precise
concentration data. This is certainly true if an element concentration in an analytical point is below 1
wt%. Fortunately, elements are often partitioned in minerais or associated with certain minerai textures.
This makes it sometimes possible to obtain chemical information even if the element's concentration in
the bulk sample is in the ppm range. The millions of micro-analyses obtained from the investigated iron
ore sample were automatically scanned with respect to EDX spectra that contain phosphorus, even in
very small amounts. Figure 10 shows an EDX spectrum of a grain of goethite obtained during
QEMSCAN® analyses compared to an ideal spectrum of goethite, which contains aluminium, silicon„
and phosphorus. The similarity of the spectra allows the conclusion that elements present in low
concentrations in the bulle sample can be traced in individual minerais. It must be assured, though, that
no peak interferences, which are generated by other elements, occur in such an EDX spectrum. Light
elements with an atomic weight <24 cause quantification problems. Concentrations of oxygen (atomic
weight 16), for example, are often underestimated in these analyses.

14
 Fe

Fe

2 1 3 le
E.M9? Aue
.1 11 14 13 13 17 29

51 otr.»»01. 3210. I. ûr. Fellari

Figure 10. EDX spectra of aluminous and siliceous goethite containing traces of phosphorus. The upper
(rad) specfnuu nus acquired in 2 milliseconds during QEMSCAN® analyses. The louer (black) spectrum
represents a modelling of an ideal goethite spectrum containing 53 wt% Fe, 4 wt% P, 5 wt% Al, 2 wt% Si.
and 36 wt% oxygen.

The investigations that led to the compilation of this paper have shown that traces of phosphorus in
aluminous and siliceous goethite are widespread and exceed in total the phosphorous content hosted
in apatite. This is shown in the QEMSCAN® images presented in Figures 11 and 12.

The absence of Ca in most of the analysed goethite spectra is also important, since it leads to the
conclusion that apatite is not the unique host minerai for phosphorus in the invesiigated iron ore.

itip4iPP
1449

Figure il. The image on the right shows glose inicroanalyses Huit contain ininor amounts of phosphorus (blue
colours). Phosphorus is concentrated in sphex ical particles (oolites).

The partitioning of phosphorus between goethite and apatite is presented in Figure 12. It appears that
goethite is the principal hast of phosphorus.

15
 ffl Goethite (P)
ffl Apatite
ti
o 0,01 0,02 0,03
Mass P

Figure 12. Partawning of phosphorus between goethate and apatzte.

SUMMARY AND CONCLUSION

The example presented in this paper demonstrates the importance of automated mineralogy for the
extraction of critical information for minerai processing of a haem.atite-rich iron ore sample. Based on
detailed mineralogical and microtextural analyses, it is possible to siraulate how physical minerai
separation can increase the quality of the ore. This requires an increase in the iron content and a decrease
in the phosphorus and sulphur concentrations. By using a simulated 4.0 g/ cra3 cut-off density, the iron
content can be increased from 45 wt% to 66 wt% at a loss of 41% of the mass, which contains 15 wt%
poorly liberated iron. The phosphorus content decreases from 0.11 wt% to 0.05 wt% in the iron-rich
fraction, white the sulphur content increases from. 1.10 wt% to 2.01 wt%. Ibis is due to the fact that
sulphur-bearing pyrite and pyrrhotite also report to this fraction. Granula-chemistry naodelling lias
shown that at a cut-off at 340 }.ira the sulphur content can be reduced to 1.04 wt% in the >340 p.in size
fraction, which represents 17 wt% of the iron concentrate. The sulphur content can be further decreased
by magnetic removal of pyrrhotite. However, the simulated result of 0.5 wt% sulphur will be rather
difficult to achieve in real processing as this requires an efficient reraoval of 1 mass% of pyrrhotite.

Similar process naodelling can be performed on any ore type to ascertain the potential for improving
grades and recoveries of economically valuable elements by using the differences in chemical and
physical properties of particles and minerais in particles. It must be kept in raind that proposed process
options are theoretical and even som.etimes hypothetical. Any proposed process option suggested by
process modelling needs to be verified by physical tests, which are performed under real conditions
that differ from ideal and computed raodelling tests. The interactions of particles with different chemical
and physical properties during minerai processing are important, difficult to simulate, and may lead to
results that do not correspond with the raodelling results. Nevertheless, m.odelling provides a roadmap
to guide process engineers in performing their tasks. In the past, and even still nowadays, minerai
processing tests were performed on almost a trial-and-error basis. Long testing cycles, accompanied by
chemical analyses, which are not always imraediately available, consume tirae and energy. This
situation can be vastly improved by using automated mineralogy and geometallurgy concepts.

The application of automated mineralogy and process raodelling is not only valuable for process
engineers. Exploration, mine planning, pre-concentration of ore, and also waste recovery can benefit
greatly from a geometallurgical approach. An ore grade evaluation based on chemical analyses, as
performed conventionally during geological surveys and mine planning, can be iraproved by
ascertaining the process behaviour of the ore well ahead of mining. Ore pre-concentration using
relatively low energy-cortsuming physical process options may upgrade a low-grade material to a
higher grade. The final decision remains in the hands of ore resource and reserve evaluators and process
engineers.

16
REFERENCES

Becker, M., Wightman, E.M., and Evans, C.L. (eds.), (2016). Process Mineraiogy. JK1122ZCMonagraph Series
in Mining and Minerai Processing, no. 6. Julius Kruttschnitt Minera]. Research Centre, Brisbane,
Australia. 470 pp.

Braun, W. (2014). Ore characterization, process mineralogy and lab automation, a road map for the
future. Minerais Engineering, (60), 63-73.

Gu, Y., Schouwstra, R.P., and Rule, C. (2014). The value of automated mineralogy. Minerais Engineering,
58. 100-103.

Lotter, N.O., Kormos, L.J., Oliveira, j., Fragomeni, D., and Whiteman E. (2011). Modern process
mineralogy: two case studies. Minerais Engineering, 24, 638-650.

Schouwstra„ R.P. and Smit, A.J. (2011). Developments in inineralogical technique - what about
raineralogists? Minerais Engineering, 24, 1124-1228.

Zhou, J. and Gu„ Y. (2016). Geornetallurgical characterization and automated mineralogy of gold ores.
Goid Ore Processing, (2nd edition). Adams, M. (ed.). pp. 95-111.

17