Journal of Minerals & Materials Characterization & Engineering, Vol. 1, No.1, pp1-9, 2002
Printed in the USA. All rights reserved
Henrique Kahn
Eliana S.Mano
and Maria Manuela M. L. Tassinari
Univ. of Sao Paulo, Polytechnic School, Dept. of. Mining and Petroleum
Engineering -
Av. Prof. Mello Moraes, 2.373 - CEP 05508-900 - Phone: +5511 3818-5787 - FAX:
+5511 3815-5785
The image analysis coupled with Energy Dispersive Spectrometry
(EDS) analysis on a Scanning Electron Microscope (SEM) was applied to the
characterization of a Zn and Pb ore as an efficient tool for mineral
quantification and for the assessment of the degree of mineral liberation.
Screened size fractions of mineral particles were mounted on polished
thin sections and, later on, analyzed by a SEM coupled with an EDS. This is
important for the discrimination of different mineral phases, particularly when
two minerals have similar average atomic numbers, such as willemite and
pyrite. Two images per field were collected - backscattered electrons image
and a multi-element X-ray dot mapping images. The results were utilized for
quantification and liberation analysis of two valuable minerals in the ore -
sphalerite and galena.
Keywords: Image analysis, mineral quantification, liberation degree analysis
Henrique Kahn
Eliana S.Mano and Maria Manuela M. L. Tassinari Vol.1, No.1
Mineral quantification and liberation degree (intergrowth) evaluations are a
routine and a major issue on ore characterization for mineral processing. Liberation
analysis is an essential data for mineral processing that shows the relative amount of
individualized grains that derives from the valuable mineral phase (“free particles’)
available for physical concentration. These analyses can be performed manually by
optical microscopy (OM) or scanning electron microscopy (SEM) in a very tiresome
and exhaustive routine. Image analysis coupled with an OM or a SEM can perform
these analyses resulting in more reliable and rapid outcomes (1, 2, 3).
Since phase differentiation by OM coupled to an image analysis is not a usual
and easy task, a digital SEM image is frequently used to solve more complex
mineralogical associations. Special care must be taken regarding sample preparation
and beam control (2,3). Atomic number contrast from backscattered electrons (BSE)
signal are primarily used for phase discrimination; however, when phases with a very
similar average atomic number are present, X-ray information is the only possible tool
that could be used to differentiate them.
This work presents an off-line image analysis routine applied to the
characterization of a partially weathered Zn-Pb ore. Six major mineral phases are
presented and must be quantified as well as the liberation analysis, which must be
performed in two valuable mineral phases - sphalerite, ZnS, and galena, PbS. The BSE
image shows clearly six distinguishable gray levels, however one of them is related to
two mineral phases with close atomic numbers - pyrite, FeS
and willemite, Zn
that could not be discriminated. These two particular phases could only be properly
segmented by coupling additional information related to their chemical composition
using X-ray data. Multi-element X-ray dot-mapping images acquired by an energy
dispersive spectrometry (EDS) were considered for this purpose.
Materials and Methods
The study samples consisted of mineral particles from four closely screened
fraction sizes mounted on polished thin sections. Special care was taken regarding the
sample preparation to avoid the physical touch of particles as well as regarding the
polishing surface quality.
BSE and X-ray dot images were obtained by a S440, Leo, coupled with an Isis-
300 EDS System, Oxford. X-ray dot-mapping images for up to 7 different elements
were acquired by S440; each selected element was represented by a binary plane and
by a specific gray level value. Both images, presenting 1024 by 768 pixels resolution,
were processed off-line by Quantimet Qwin-Pro software, Leica, an image analysis
system which operates under the same SEM PC hardware.
In order to determine all the mineral phases, qualitative mineralogical work
was first performed coupling X-ray diffraction data with a detailed SEM-EDS
observation. Fig. 1 shows the BSE image of the particulate material of the size fraction
between 20 and 37µm. The major mineral phases are constituted essentially by
sphalerite, galena, pyrite, willemite, quartz and dolomite; barite, apatite and ilmenite
occur as trace minerals (<1%) and were not discriminated by image analysis (others).
Fig. 1 - Particulate material of a size fraction mounted in
a thin-polished section presenting the major mineral phases.
The second step comprised the acquisition of BSE and X-ray dot-mapping
images. Since the acquisition time for the dot images were relatively high, off-line
image processing was chosen to assure a better SEM electron beam stability during the
total acquisition period, which corresponds to almost 200 minutes for 30 fields per
sample. Incident probe current, brightness and contrast levels were set to allow the
acquisition of BSE and X-ray images with a good quality for further processing.
Because particle density per field is one of the major factors that directly affect
the total processing time, an ideal compromise is required to optimize the acquisition
time. The SEM magnification was adjusted for an average of 40 to 50 particles per
field, a situation in which some particles may touch other particles. For this reason, the
first step in image processing is to individualize these touching particles.
A relatively complex subroutine (2), summarized in the Fig. 2, was applied to
discriminate the touching particles. Firstly, the detected image (Fig. 2a) was eroded
(Fig. 2b), and then skeleton and prune operations were applied in order to separate the
particles so they would not touch each other. Finally applying outline followed by
close and open operations to the particles resulted in the dark lines of potential
touching areas (Fig. 2d). These lines were subtracted from detected image (Fig. 2a) by
logical operation, resulting in the final binary image of particles to be measured.
Henrique Kahn
Eliana S.Mano and Maria Manuela M. L. Tassinari Vol.1, No.1
Fig. 2 Separation of touching particles.
An image analysis routine was developed in order to discriminate the mineral
phases and, later on, to perform modal and mineral liberation analysis. Detection,
identification and segmentation of the phases are the most complex issues, and the
routine should process a gray scale image plus external inputs as the X-ray dot image.
Gray level threshold from the BSE image allowed discriminating up to 6 binary
planes - galene, sphalerite, “pyrite+willemite”, hematite, quartz and dolomite (Fig. 3).
Fig. 3 Superimposed binary images of the detected phases after processing
the BSE image. Pyrite and willemite, due to their close atomic numbers and
complex intergrowths, were included in the same binary plane, as shown in
Fig. 4.
2a 2b
Fig. 4: Binary plane image of pyrite plus willemite
The acquired multi-element X-ray dot-mapping image was then submitted to a
gray level threshold that was intended to discriminate the sulfur and zinc individual X-
ray binary images as shown in the Fig. 5.
Fig. 5: Sulfur and zinc X-ray dot-mapping binary images
after processing the multi-element X-ray dot image
Logical operations with the “pyrite+willemite” binary image as a mask (Fig. 4)
and X-ray dot images from S and Zn, Fig. 5, resulted in two others images representing
S from pyrite and Zn from willemite, as showed in Fig. 6.
Fig. 6: Sulfur from pyrite and zinc from willemite after
logical operations between “pyrite+willemite” binary image
and X-ray dot images from S and Zn
Henrique Kahn
Eliana S.Mano and Maria Manuela M. L. Tassinari Vol.1, No.1
Further closing and logical operations allowed the segmentation between these
two phases, as observed in Fig. 7.
Fig. 7: Results of the final segmentation between pyrite and
willemite after processing the gray level and X-ray dot-
mapping images
At the end of the segmentation procedure each mineral phase was represented
by a binary image plane. Modal or quantitative phase analysis could be then performed
considering the area fraction measurements for the different binary planes (mineral
phase). The results of all the 30 fields were accumulated in a file and, later on,
normalized to 100% regarding the volume percentage. The weight percentages were
attained considering the mineral densities and their volume fractions.
Liberation analyses of sphalerite and galena were performed by the liberation
spectra technique. Feature measurements and a coincident parameter subroutine
allowed the determination of the total area and the area fraction for each particle
containing the valuable minerals. The particles containing sphalerite or galena were
each subsequently and individually classified according to its area fraction and the
results were expressed as cumulative curves compared to the amount of the valuable
mineral phase (liberation spectra or potential recovery curves).
This procedure is not only the most suitable for the establishment of flexible
criteria of free particles, but also gives important information regarding the distribution
of locked particles. Stereological corrections for the liberation data from 2D to 3D
were not applied to the attained results.
The methodology above was applied to four closely screened size fractions.
All the major mineral phases could be distinguished and quantified as the example
exposed in Fig. 8 and by the results presented in Table 1.
Fig. 8: Segmented mineral phases after image analysis
Table 1: Modal analysis results (weight percentage)
Mineral Size fraction
Phase +44µm -44 +37µm
-20 +10µm
sphalerite 4.3 4.3 3.8 3.4
galena 1.1 1.2 1.8 2.7
pyrite 2.2 3.2 3.3 3.5
willemite 2.6 2.3 2.5 1.7
hematite 9.1 8.6 8.5 9.2
dolomite 76.7 77 77 77
quartz 4.0 3.6 3.3 2.6
others <1.0 <1.0 <1.0 <1.0
Sphalerite, the major Zn bearing mineral, represents 3.8% in weigh and carries
about 60% of the total zinc of the sample. EDS analysis showed minor contents of Fe
and Cd associated with sphalerite, with an average grade of 0.6% in weigh for both
elements. The average grain size of the sphalerite was between 10 and 20µm.
Willemite, 2.1% in weight, is an autigenic Zn bearing mineral related to the weathering
process that carries almost 30% of the zinc content in the sample.
Galena, 1.8% in weight, is the only Pb bearing mineral. Two distinct
generations could be identified; the first and coarsest one presents an average grain
size of 30 to 40µm. The other one, with grain sizes below 10µm, is usually intergrown
with sphalerite, willemite but seldom with dolomite.
Dolomite (and Zn-dolomite) is by far the major mineral in the sample,
constituting 76% of the total in weight. Detailed SEM study showed that dolomite is
partially recrystallized with incorporation of variable amounts of Zn and Fe. EDS
analyses showed ZnO ranging from 0.4% up to more than 10%, with an average grade
of 4.9% of ZnO, which means that 10% of the total Zn of the sample is related to
Henrique Kahn
Eliana S.Mano and Maria Manuela M. L. Tassinari Vol.1, No.1
Mineral liberation spectra, demonstrated by both sphalerite and galena, are
observed by Figs. 9 and 10, respectively.
Fig. 9: Liberation spectra curves for sphalerite
Fig. 10: Liberation spectra curves for galena
To achieve relatively high grades and recoveries in mineral processing the
valuable mineral must be individualized to almost monomineral particles by grinding
operations prior to the physical concentration process. The valuable mineral content is
directly related to the concentrate specifications, but, on average, it is acceptable for a
90% fraction of the valuable mineral to consider a particle as “free”.
For the studied sample reasonable mineral liberation is attained only below
20µm, which means that this type of ore must be ground in much a finer way than
usual (<44µm). Potential recoveries of 85-90% for galena and 75-85% for sphalerite or
45-51% for Zn can be expected.
An image analysis routine coupling BSE and X-ray dot-mapping images
produced a proper mineral discrimination for quantitative phase and mineral liberation
degree analysis. The data acquired for closely screened fractions sizes were extremely
helpful for understanding the ore behavior and the optimization of the mineral dressing
process. The study samples represented three Zn bearing minerals; sphalerite, the
valuable Zn mineral, carries only 60% of the total Zn sample content. Liberation
analysis showed that ore must be ground much finer than usual, minus 20µm, to assure
the required concentrate grades with reasonable mineral recoveries.
1. Gabas, S.G. Análise de Imagens Aplicada à Caracterização de Minérios Análise
Modal e Liberação. Dissertação de Mestrado, Escola Politécnica da Universidade de
São Paulo (1999).
2. Kahn, H.; Sant’Agostino, L.; Mano, E.S., Tassinari, M.M. (1998) Acta Microscopica.
7-A: 241-244.
3. Lastra, R.; Petruk, W; Wilson, J. Image analysis techniques and applications to mineral
processing. In: Modern approaches to ore and environmental mineralogy.
Mineralogical Association of Canada. Short Course Series, V.27, p327-366