Professional Documents
Culture Documents
Abstract — The determination of mineral liberation by microscopic examination of ore particle sections
is stereologically biased. Procedures have been developed to correct this bias, but most have not been
thoroughly tested. A standard material, a material of known particle composition and locking character-
istics, would allow the proper testing of these procedures. In this work, a method was developed for the
creation of a standard material consisting of two phases: glass and lead borate. Simple-locked particles
and particles with granular and layered textures can be created. The standard material is versatile and
easy to use.
Methods have been developed to correct the stereological For the standard material, there must be a method of
bias [1, 2, 3, 4, 5, 6, 7], but the effectiveness of these methods distinguishing the two phases. If optical microscopy is
is difficult to test since the true liberation is not known. used, the two phases must be different colours or have
Standard materials, that is, materials of known particle libera- some other differentiating feature. If electron microscopy is
tion and locking characteristics, have been developed in an used, the two phases must have different grey levels when
attempt to allow the assessment of these correction proce- viewed with backscattered electron imaging. The backscat-
dures. The present work describes a method of producing a tered grey level of a phase is directly related to its average
versatile standard material that can be used to test the robust- atomic number such that a low atomic number material will
ness of stereological correction procedures. appear dark and a high atomic number material will appear
CANADIAN METALLURGICAL QUARTERLY
122 D. LIN and J.A. FINCH
bright. In this work, the standard material was designed for porosity in the ore [6(p.118)] and there was some variation
use with electron microscopy. in the density of the iron minerals.
It is desirable that the standard material be able to 2. Bole et al. [9] examined two binary ores: an iron-
exhibit simple locking (i.e., only one interface between the oxide/silicate sample and a sphalerite/dolomite sample.
phases in the particle). Simple locking is probably the most Both ores were crushed and screened to 417-595 µm and
important of all locking types in mineral processing fol- the iron-oxide/silicate particles fractionated using heavy
lowing the argument that since comminution is aimed at liquids. The densities of the silicate and iron oxide phases
creating free particles, the mineral which fails to be liberat- were 2.72 and 5.10 g/ml, respectively. The sphalerite/dolomite
ed is most likely concentrated in the next simplest class - particles were density fractionated using the Magstream
the simple-locked particle class. Simple locking also cre- separator (a centrifugal magnetogravimetric separator).
ates the most severe stereological bias and is, therefore, a The density of the dolomite was 2.85 g/ml and that of the
severe test of any stereological correction procedure [1,2]. sphalerite, 4.00 g/ml.
The standard material should also be flexible enough to 3. Miller and Lin [10] used a unique approach to deter-
simulate different types of locking textures. Although a mining the true liberation of mineral particles. They used
stereological correction procedure may be effective with serial sectioning to determine the liberation of an iron ore
one type of locking texture, it may not be as effective with and a copper ore. Particles were screened to 74-105 µm and
others. In addition, it would be instructive to examine the mounted in resin. By polishing this pellet down by intervals
effect of locking texture on the predictive abilities of dif- of 18 µm, a series of parallel sections was gathered for each
ferent liberation models. particle. The composition of each particle was reconstruct-
ed based on these sections. Although serial sectioning does
There are two categories of standard materials: natural- provide an unbiased measure of the true liberation distrib-
ly-occurring and artificial. A short review of previous ution, it is too tedious and expensive to use in regular lib-
attempts to create standard materials follows. eration analyses.
used as a source of lead for glass melts or ceramic glazes, but 4. The power was shut off to the furnace and the crucible
in this work it was used as one of the phases of the standard was allowed to cool to room temperature.
material. The melting point of lead borate is approximately
500°C while the melting point of glass is approximately 5. The sample was removed from the crucible, crushed
800°C. The standard material was created by melting the lead and wet-screened to remove the fines.
borate and embedding the glass particles. The glass particles
softened slightly and produced a strong bond with the lead 6. The 425-600 µm fraction was screened out.
borate. (It was later determined that some glass diffused into
the borate which further strengthened the bond.) A SEM (scanning electron microscope) backscattered
electron micrograph of this material is shown in Figure 2.
Glass was used as the grain material because it has sim- A large majority of the sections exhibit locking. There were
ilar brittleness and breakage properties to lead borate. a few voids, but they were small (<10 µm) and were pre-
Silica was considered as the grain material, but the use of sent in only approximately 1% of the borate sections. It
glass rather than silica allows greater control over the lock- appeared that there were slightly more simple-locked sec-
ing texture (as seen by the creation of the layered texture tions than complex-locked sections. The complexity of the
discussed later) and because it has a lower melting point. complex-locked sections was low; most of the complex-
The density of the lead borate and glass were measured locked sections contained only two interfaces.
with a water displacement technique using a pycnometer
bottle [15] to be 6.833 g/ml (±0.006)1 and 2.502 g/ml
(±0.001), respectively. Subsequent experimentation showed
that after contact some glass diffused into the lead borate to
give a SiO2 content of 5.95% (±1.38%) as determined by
microprobe analysis. The density of the borate was re-
determined as 6.052 g/ml (±0.009) and this was the value
used in the density fractionation.
2. The slides were placed in a crucible containing borate A SEM backscattered electron micrograph of these par-
powder so that only the lower portion of the long slides was ticles is shown in Figure 4. Only simple locking was
in contact with the powder; the short slides and the clamp observed; no sections with complex locking were seen.
itself were clear of the powder.
The particle size of the simple-locked particles was
3. The crucible was placed in an induction furnace pre- selected to be 425-600 µm because the granular-textured
heated to 600°C. This temperature was maintained for 45 standard material was this size. If the simple-locked and the
minutes. granular-textured standard materials were to be combined
to form a liberation distribution, the particles would have to
4. The power was shut off to the furnace and the crucible be the same size. Simple-locked and layered-texture parti-
was allowed to cool to room temperature. cles are a useful supplement to the granular-textured stan-
dard material.
5. The clamp and the short glass slides were removed.
6. The sample was removed from the crucible, crushed Density fractionation
and wet-screened to remove the fines. To determine the liberation, the particles were separated
into different density classes using the Magstream separa-
7. The 425-600 µm fraction was screened out. tor. In this work, density fractionation was performed only
on the granular-textured standard material particles.
borate. This leaves a 1% gap in composition between frac- affects the IA grade. It is important to generate random sec-
tions, but helps minimise misplaced particles. tions. In these samples, the particles in each fraction should
have very similar densities and consequently, there should
The particles that were -5 vol.% borate or +95 vol.% be little segregation. But, regardless, precautions were
borate were considered to be free. True free glass particles taken to ensure random mounting of the particles.
(i.e., 100% glass) were created simply by crushing glass to
425-600 µm. True free borate was created by placing borate The sample preparation involved mixing 0.4 ml of the
powder alone in porcelain crucibles and heating to 600°C sample with an equal amount of a diluent material (425-600
for 30 min. After the borate had cooled and solidified, it µm graphite particles) and placing the mixture at the bot-
was removed from the crucible and crushed to 425-600 µm. tom of a 32 mm mold. The diluent helps to support the sam-
ple particles in space (and thus reduce preferential orienta-
tion and the incidence of particles touching) and promotes
Measurement of the grade of the composition fractions random orientation [25]. Graphite was selected as the dilu-
Following Magstream separation, the grade (i.e., the per- ent since it has a low density and a grey level similar to the
cent borate) of the composition fractions was determined. mounting medium resin when observed by backscattered
Two different methods were used to measure the grade: electron imaging. The mounting medium was Epofix resin
water displacement in a pycnometer bottle and image (Struers Inc.). The resin was prepared by mixing with the
analysis. Henceforth, the grade measured using these meth- hardener and then centrifuging at 1600 rpm for 1 minute to
ods will be referred to as the pycnometer and IA grades, remove any air bubbles. A small quantity of the resin (0.8
respectively. ml) was added to the particles in the mold and mixed. The
small amount of resin helps inhibit preferential settling by
The IA grade is determined by microscopic examina- restricting the space in which the particles can move. The
tion of the particle sections. The total area of the two phas- mold was then centrifuged at 1600 rpm for 5 minutes to
es is measured with an image analyser. It is known that the remove any air bubbles that may have resulted from the
area grade is not stereologically biased and is identical to mixing. Afterwards, more resin was added to top off the
the volumetric grade [6(p.22)]. mold.
In sample preparation, prevention of preferential set- For the standard material, the pycnometer and IA grades
tling (due to density differences between the particles) and were measured for each composition fraction. The two
preferential orientation (due to density differences between grades should be close to the midpoint. The results are
the phases in the locked particles) is important since this shown in Table I.
Table I — Pycnometer and IA grades of the composition fractions of the standard material (density of borate,
6.052 g/ml; glass, 2.502 g/ml)
For all the fractions, both the pycnometer and IA grades REFERENCES
fall in the composition fraction range and in most cases,
they are close to the midpoint of the range and the differ- 1. G.S. Hill. Applications of two-dimensional image analysis to
ences in grades are not significant. mineral liberation studies. Ph.D. thesis, McGill University
(1990).
The pycnometer grade of the +95 vol.% borate fraction
2. G.S. Hill, N. Rowlands and J.A. Finch. Data correction in
was greater than 100 vol.%. This can be explained by con- two-dimensional liberation studies. Process Mineralogy VII,
sidering the creation of the standard material. Although the A. Vassiliou, D. Hausen and D. Carbon eds. (Warrendale,
glass is always in contact with the borate, there is a layer of PA: The Metallurgical Society of AIME), 617-32 (1987).
borate at the bottom of the crucible (the glass grains float in
the borate due to their lower density) that does not come in 3. C.L. Lin, J.D. Miller and J.A. Herbst. Solutions to the trans
contact with the glass. The borate density was calculated formation equation for volumetric grade distribution from
only for borate that had been in close contact with glass linear and/or areal grade distributions. Powder Technol., 50,
grains and where limited diffusion had occurred. All the 55-63 (1989).
high density borate reports to the +95 vol.% borate fraction
where it belongs. 4. C.L. Schneider, C.L. Lin, R.P. King and J.D. Miller. Improved
transformation technique for the prediction of liberation by
a random fracture model. Powder Technol., 67, 103-11 (1991).
Use of the standard material 5. G. Barbery. Mineral liberation: measurements, simulation
The standard material can be used to produce different lib- and practical use in mineral processing (Quebec: Les Editions
eration distributions by combining the different composi- GB), 351p. (1991).
tion fractions. These distributions can be used as a check on 6. S.L. Gay. Liberation modelling using particle sections. Ph.D.
the effectiveness of stereological correction procedures thesis, University of Queensland (1996).
provided the other components of liberation analysis - sam-
ple preparation and image processing - are performed cor- 7. D. Lin. A standard material for liberation analyses and exam-
rectly. The known true liberation distribution can be com- ination of the robustness of stereological correction procedures.
pared with the corrected distribution. We have reported on Ph.D. thesis, McGill University (1997).
this comparison elsewhere [26]. In more general terms, the
standard material would enable the whole procedure (not 8. P.S.B. Stewart and M.P. Jones. Determining the amounts and
just stereological correction) to be tested, although the the compositions of composite (middling) particles. Proc.
source of error may be difficult to isolate. In this regard, a 12th Int. Mineral Process. Congress, Sao Paulo, Meeting B,
paper 4 (1977).
standard material is superior to using computer-simulated
particles since it involves all the steps of a liberation analy- 9. J. Bole, C.L. Lin and J.D. Miller. Experimental verification of
sis. Since the two phases of the standard material are the PARGEN simulator for liberation analysis. Int. J. Mineral
strongly bonded to each other, the standard material can Process., 37, 209-21 (1993).
also be used to test liberation models. Most of these mod-
els assume negligible breakage along grain boundaries. 10. J.D. Miller and C.L. Lin. Treatment of polished section data
for detailed liberation analysis. Int. J. Mineral Process., 22,
41-58 (1988).
CONCLUSIONS
11. P.S. Bagga. Simulation of the liberation phenomena in mineral
systems. Ph.D. thesis, Pennsylvania State University (1983).
A method was developed for the creation of a standard
material consisting of glass and lead borate. This standard 12. L.C. Woollacott and M. Valenta. Use of synthetic ore particles
material is flexible and easy to use. Particles with granular to test a transformation function in liberation analysis. Minerals
and layered textures were created along with particles Engineering, 9(10), 1017-32 (1996).
exhibiting simple locking. The standard material can be
used to test the steps in liberation analysis including stere- 13. D. Lin. The production of a standard material for liberation
ological correction procedures. analysis. M.Eng. thesis, McGill University (1991).
18. M.S. Walker, A.L. Devernoe, W.S. Urbanski and G.R. Morrow.
Mainly "gravity" separations using magnetic fluids under
rotation. Proc. XVII Int. Mineral Process. Congress, Dresden,
105-16 (1991).