XXXI International Mineral Processing Congress 2024 Proceedings/Washington, DC/Sep 29–Oct 3 1579
interaction spot and an acquisition speed of 10 ms per
pixel. Finally, a lateral resolution of 40 μm was employed.
Energy Dispersive X-ray fluorescence (ED-XRF).
The samples were pulverized to produce particles with a
diameter of approximately 20–30 µm. The sample cups
were surrounded by a ~4 µm polypropylene film to opti-
mize the detection of X-rays for low energy elements such
as Al, Na, K, etc. The aliquots were analyzed by X-ray fluo-
rescence using a handheld Niton XL3t GOLDD+ spec-
trometer (Thermo Scientific, United States). The machine
is run with an Ag anode X-ray tube working with a voltage
of 50 kV and it is equipped with a silicon drift detector
(SDD). One analysis runs for 120 seconds and is divided
into four filters (Main, Low, High, and Light). The four
acquired spectra are automatically interpreted by the min-
ing mode Hf/Ta. Each sample underwent 3 measurements
at the same location.
Laser Ablation
Laser Ablation and Inductively Coupled Plasma Mass
Spectrometry. LA-ICP-MS measurements were con-
ducted using a 193 nm wavelength nanosecond Excimer
laser (GEOLAS Pro). This laser enables sample ablation
placed within an ablation cell. A helium flux transports the
ablation products to an Agilent 7500 inductively coupled
plasma mass spectrometer (ICP-MS) for analysis. The anal-
ysis range includes all elements from lithium to uranium
with a detection limit on the order of ppm. Analyses were
performed on the polished section containing the larg-
est grains (10–500 µm) with a 20 µm spot. In total, 45
individual analyses were conducted on lepidolite grains to
determine the lithium content of this mineral.
Micro laser-induced breakdown spectroscopy
(µLIBS). LIBS mapping measurements were conducted
at the Institut Lumière Matière laboratory (Lyon, France).
The procedural approach followed was that developed by
Cáceres and colleagues. LIBS mapping was performed
using a homemade optical microscope and a Nd:YAG
laser (Centurion GRM, Quantel by Lumibird) with an 8
ns pulse duration operating at 100 Hz. The plasma emis-
sion was collected by three lens-fiber systems coupled to
two spectrometers to simultaneously detect strong emission
lines from all the elements of interest. Samples were placed
onto an XYZ stage and scanned pixel by pixel with a lat-
eral resolution of 20 μm. Further details on the procedural
method can be found in publications (Cáceres et al., 2017
Richiero et al., 2022).
Python codes
MAsking spectRosCopIc dAtacube (MARCIA)
The MARCIA code was developed at the GeoRessources
laboratory (France). The objective of this Python code is to
perform manual classification of hyperspectral data from
µXRF or BSE/EDS/SEM. This is achieved by defining
masks, which are linear combinations of elemental intensi-
ties from the obtained spectra (or colors in the case of BSE
images). Once the masks, representing mineral classes, are
defined, they can be used in analysis equipment software
for further data processing. More details about the code
and mask classification can be found on GitHub (10.5281/
zenodo.3929744).
Modal Mineralogy Matrix Calculation (EMC—
Element to Mineral Conversion)
Matrix calculations for reconstructing modal mineralogy
from chemical assays are widely used in the field of geology.
They are defined as follows:
e C m
c11
c
c1
c
m1
m
e1
e
n1
n
nn n n
+##h
g
j
g
h h hH ==H H (1)
where C is the matrix of chemical composition of differ-
ent mineral phases, m is the vector containing the modal
mineralogy of the sample (unknown from the equation),
and e is the vector containing the chemical composition of
the sample.
Usually, C is determined using an electron probe micro-
analyzer (EPMA), while e is determined using spectro-
scopic methods such as XRF or ICP. In our case, since the
lithium content is not obtained by this analytical method,
EDS and laser ablation measurements were combined to
obtain the total chemistry of mineral phases. The method
for solving this matrix calculation is adapted from the
Tolosana-Delgado and co-workers method and was imple-
mented using Python programming language utilizing the
Quadratic Programming package (Tolosana-Delgado et al.,
2011).
Beneficiation
Sample Preparation
The provided cores were crushed using a jaw crusher, a gyra-
tory crusher, and a roll crusher. The resulting sample was
then ground using a laboratory-scale ball mill. Grinding
tests were conducted with 1 kg of ore, 10 kg of balls, and
500 mL of water (66% solids). The fines (10µm) were sep-
arated using a 1-inch hydrocyclone (Mozley). This sample
is the feed for flotation tests.
interaction spot and an acquisition speed of 10 ms per
pixel. Finally, a lateral resolution of 40 μm was employed.
Energy Dispersive X-ray fluorescence (ED-XRF).
The samples were pulverized to produce particles with a
diameter of approximately 20–30 µm. The sample cups
were surrounded by a ~4 µm polypropylene film to opti-
mize the detection of X-rays for low energy elements such
as Al, Na, K, etc. The aliquots were analyzed by X-ray fluo-
rescence using a handheld Niton XL3t GOLDD+ spec-
trometer (Thermo Scientific, United States). The machine
is run with an Ag anode X-ray tube working with a voltage
of 50 kV and it is equipped with a silicon drift detector
(SDD). One analysis runs for 120 seconds and is divided
into four filters (Main, Low, High, and Light). The four
acquired spectra are automatically interpreted by the min-
ing mode Hf/Ta. Each sample underwent 3 measurements
at the same location.
Laser Ablation
Laser Ablation and Inductively Coupled Plasma Mass
Spectrometry. LA-ICP-MS measurements were con-
ducted using a 193 nm wavelength nanosecond Excimer
laser (GEOLAS Pro). This laser enables sample ablation
placed within an ablation cell. A helium flux transports the
ablation products to an Agilent 7500 inductively coupled
plasma mass spectrometer (ICP-MS) for analysis. The anal-
ysis range includes all elements from lithium to uranium
with a detection limit on the order of ppm. Analyses were
performed on the polished section containing the larg-
est grains (10–500 µm) with a 20 µm spot. In total, 45
individual analyses were conducted on lepidolite grains to
determine the lithium content of this mineral.
Micro laser-induced breakdown spectroscopy
(µLIBS). LIBS mapping measurements were conducted
at the Institut Lumière Matière laboratory (Lyon, France).
The procedural approach followed was that developed by
Cáceres and colleagues. LIBS mapping was performed
using a homemade optical microscope and a Nd:YAG
laser (Centurion GRM, Quantel by Lumibird) with an 8
ns pulse duration operating at 100 Hz. The plasma emis-
sion was collected by three lens-fiber systems coupled to
two spectrometers to simultaneously detect strong emission
lines from all the elements of interest. Samples were placed
onto an XYZ stage and scanned pixel by pixel with a lat-
eral resolution of 20 μm. Further details on the procedural
method can be found in publications (Cáceres et al., 2017
Richiero et al., 2022).
Python codes
MAsking spectRosCopIc dAtacube (MARCIA)
The MARCIA code was developed at the GeoRessources
laboratory (France). The objective of this Python code is to
perform manual classification of hyperspectral data from
µXRF or BSE/EDS/SEM. This is achieved by defining
masks, which are linear combinations of elemental intensi-
ties from the obtained spectra (or colors in the case of BSE
images). Once the masks, representing mineral classes, are
defined, they can be used in analysis equipment software
for further data processing. More details about the code
and mask classification can be found on GitHub (10.5281/
zenodo.3929744).
Modal Mineralogy Matrix Calculation (EMC—
Element to Mineral Conversion)
Matrix calculations for reconstructing modal mineralogy
from chemical assays are widely used in the field of geology.
They are defined as follows:
e C m
c11
c
c1
c
m1
m
e1
e
n1
n
nn n n
+##h
g
j
g
h h hH ==H H (1)
where C is the matrix of chemical composition of differ-
ent mineral phases, m is the vector containing the modal
mineralogy of the sample (unknown from the equation),
and e is the vector containing the chemical composition of
the sample.
Usually, C is determined using an electron probe micro-
analyzer (EPMA), while e is determined using spectro-
scopic methods such as XRF or ICP. In our case, since the
lithium content is not obtained by this analytical method,
EDS and laser ablation measurements were combined to
obtain the total chemistry of mineral phases. The method
for solving this matrix calculation is adapted from the
Tolosana-Delgado and co-workers method and was imple-
mented using Python programming language utilizing the
Quadratic Programming package (Tolosana-Delgado et al.,
2011).
Beneficiation
Sample Preparation
The provided cores were crushed using a jaw crusher, a gyra-
tory crusher, and a roll crusher. The resulting sample was
then ground using a laboratory-scale ball mill. Grinding
tests were conducted with 1 kg of ore, 10 kg of balls, and
500 mL of water (66% solids). The fines (10µm) were sep-
arated using a 1-inch hydrocyclone (Mozley). This sample
is the feed for flotation tests.