Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

1 s2.0 S0892687521003198 Main

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Minerals Engineering 171 (2021) 107090

Contents lists available at ScienceDirect

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

Understanding gangue acid consumption in copper sulfide heap leaching:


Predicting the impact of carbonates, silicates and secondary precipitates
Mark Thomas
Geometallurgy Consultant, Ore Analytics Limited, 8 Thornberry Terrace, Penzance, Cornwall TR18 3AH, UK

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

Keywords: A geometallurgical model, linked to simple ore body characteristics, was established to understand short- and
Geometallurgy long-term gangue acid consumption (GAC) variability for a new copper sulfide chemical heap leaching process
Modelling under standard metallurgical conditions, including an acid dose of 12 kg/t sulfuric acid during agglomeration at
Copper
a particle size of 80% passing ¾”. Carbonate availability, or liberation, defined as a function of feed grade and
Heap leaching
Gangue acid consumption
alteration type, enabled the calculation of short-term GAC when combined with mineral abundance, assuming
stoichiometric behavior. An adjustment to the acid balance needed to be applied for jarosite formation, which
was made possible via a linear correlation, again related to carbonate feed content. The long-term GAC
component of the ore could then be determined by difference, comparing data from the column program to the
estimation of the two previous variables, resulting in the identification of biotite and chlorite as key drivers of
long-term acid consumption during leaching. Final inputs for the three-stage approach to predict overall GAC
were simplified to carbonate (CO3), magnesium (Mg), and iron (Fe), facilitating implementation in the 3D
resource block model. Application of this methodology indicated that the magnitude of long-term GAC is reduced
when higher levels of acid are consumed during the initial phases of processing. The study clearly shows how a
holistic knowledge of the ore body can provide an in-depth understanding of future processing response,
enabling key investment decisions to be made appropriately, prior to mine construction.

1. Introduction reagents, like ferric sulphate in bioleaching (Petersen, 2016), or chloride


e.g., CuproChlor (Montealegre, 2002) a process developed at Michilla in
In porphyry copper heap leaching systems, an interest in gangue Chile. While copper extraction can often be improved by increasing the
mineralogy should be of primary importance, obviously in addition to acid dose, particularly in oxide leach where greater solubility is expe­
the copper-bearing minerals, since this can significantly impact the rienced at lower pH, this tends to augment gangue acid consumption as
overall metal recovery, as well as operational costs (Chetty, 2018). This well (Lu et al., 2017).
process involves crushing the ore, to a particle size of 80% passing (P80) The low-grade nature of porphyry copper mineralization (Sinclair,
5 to 100 mm (John, 2011), but usually below 25 mm or 1′′ , with most 2007) means that silicates, or non-sulfide gangue minerals, represent
operations employing agglomeration, where acid, water and possibly most of the feed, implying that due consideration should be given to how
other reagents are added, prior to heaps being constructed in lifts from 6 this will react during the long cycles associated with heap leaching
to 10 m (Petersen, 2016). Most heap leach operations consider 80% of (Chetty, 2018). Certain mineralogical characteristics, such as finer
gangue acid consumption (GAC) to be added, prior to the curing stage, gangue mineral grain size, result in increased reactivity and can there­
whilst the other 20% is utilized during leaching (Salgado et al., 2009). fore increase gangue acid consumption, whereas acid regeneration may
The irrigation solution used during leaching, known as raffinate, is occur during precipitate formation e.g., jarosite or alunite (Jansen and
sulfuric acid-based in the copper industry. Oxides can be processed Taylor, 2003). Production of these precipitates reduce the degree of
under reasonably mild acidic conditions assuming pH < 5 (Schlesinger gangue acid consumption since the release of acid alters the overall
et al. 2011), because most of the target minerals in this ore are highly balance (Free, 2010). Pyrite oxidation in bioleaching also generates
soluble (Ghorbani et al., 2015), although copper sulfides are relatively sulfuric acid (Nowaczky and Domka, 2000). Finer particle size distri­
more refractory and require the addition of aggressive oxidizing butions as well as other relevant metallurgical parameters, including

E-mail address: mark.thomas@oreanalytics.co.uk.

https://doi.org/10.1016/j.mineng.2021.107090
Received 28 March 2021; Received in revised form 28 June 2021; Accepted 21 July 2021
Available online 14 August 2021
0892-6875/© 2021 Published by Elsevier Ltd.
M. Thomas Minerals Engineering 171 (2021) 107090

higher in-heap temperatures or greater acid addition/concentration, for measure the variability of metallurgical response, as a function of
instance, drive increases in GAC, however, taller heaps or shorter leach mineralogical composition, such as an initial acid dose of 12 kg/t sul­
cycles can mitigate this somewhat (Jansen and Taylor, 2003). furic acid for curing. The specifics of these conditions are not presented
Gangue acid consumption generally occurs via 3 mechanisms: (a) here due to confidentiality reasons; however, the focus of this paper is
mineral dissolution; (b) initial dissolution followed by the formation of a related to the data collection, analysis, and modelling methodology used
precipitate layer restricting further leaching; or (c) transformation of to produce an understanding of the spatial variability in gangue acid
gangue into secondary species (Free, 2010). consumption, before mining and processing. Monitoring of raffinate and
Pregnant Leach Solution (PLS) occurred daily supplying information on
• Carbonates react quickly in acid (Van Staden et al., 2009) and impurity concentrations and acid consumption over time. At the end of
consequently have the potential to cause serious impacts in the the cycle, columns were unloaded, and metallurgical balances closed out
curing stage where the ore is left to rest, usually for a period of days with solids analysis.
or weeks, after heap construction. Examples include a decrease in pH QEMSCAN (Quantitative Evaluation of Minerals by SCANning elec­
and subsequent precipitation of secondary minerals. Also, failure to tron microscopy) generated modal mineralogy for the project via Bulk
add sufficient acid, during the agglomeration process, may result in Mineralogical Analysis (BMA), at a line spacing of 100 μm, a pixel
complete reagent consumption prior to irrigation; lower metal re­ spacing of 3 μm, and a count rate of 1000 counts per pixel. Particle
covery commonly follows such circumstances (John, 2011). Mineralogical Analysis mode (PMA) was also used to collect additional
• Calcite dissolution is faster than dolomite, which is in turn faster than textural information. Three carbon-coated epoxy-resin mounted blocks
that of siderite, although kinetics vary as a function of pH, solution were fabricated per sample, using 1-gram aliquots at P80 500 μm, and a
composition or temperature, as well as morphological features like graphite to sample ratio of 2 to 1. Drill core were initially crushed to
grain size, surface area and the distribution of reactive surface sites 95% passing (P95) 10# Ty, 500 g subsamples were collected after ho­
(Morera-Chavarria et al., 2016). mogenization, then further crushed to P95 600 μm, and final aliquots
• The amount of acid consumed by carbonates is finite, however, so supplied via use of a rotary sample divider for mineralogical prepara­
effectively the reaction stops once all the available mineral has tion. A top size at least 10 times that the texture of interest was selected
dissolved. (Jones, 1987), with protocol design centred around copper sulphides, to
assure a fundamental sampling error (FSE) for copper under 5% at 95%
The reactivity of silicate gangue tends to occur during the leaching confidence, based on the average sampling constant (C) of the deposit.
stage, with acid consumption being most significant prior to this during This is taken as an acceptable level of error whilst preserving sufficient
irrigation. Biotite dissolution is generally constant (Acker and Bricker, ore texture for meaningful quantitative analysis. Sutherland (2007)
1992) whilst incongruent leaching behavior has been reported for suggests a minimum of 200 measurements on the mineral of interest to
different chlorite species (Tan et al., 2012). Iron-rich chlorites, like define grain size. In this case, 750 to 1,500 particles were collected
chamosite, appear to be more soluble, resulting in relatively higher acid during PMA, and 60,000 to 100,000 intercepts (BMA), per sample, using
consumption compared to magnesium-rich chlorites (Jansen and Taylor, standard measurement parameters, so that a robust assessment of key
2003). To provide context, in terms of the rates of acid consumption, gangue mineralogical attributes could be undertaken.
other studies have shown that calcite exhibits much more rapid disso­ Quartz, copper, and gold reference standards facilitated QEMSCAN
lution than that of biotite and chlorite, however, these minerals system calibration; data processing utilized iDiscover software applying
demonstrate kinetics several orders of magnitude faster than feldspars a deposit-specific mineral reference library or SIP file (Species Identifi­
like albite and orthoclase (Holland and Turekian, 2014). cation Protocol). Independent analysis of copper, iron, sulfur, and sili­
Consequently, an understanding of mineralogy is crucial to accu­ con formed the basis of quality control, where the objective was for
rately predict gangue acid consumption as each species has its own calculated assays, as a function of mineralogy, to be within ± 20% of
unique dissolution kinetics and solubility (Chetty, 2018). Application of measured chemical values for each target element. Inductively Coupled
fixed or assigned performance indices to estimate GAC (e.g., by ore type) Plasma Mass Spectrometry (ICP-MS) enabled quantification of 40 major,
for valuation, or other planning purposes, is risky as this can change minor and trace elements, following a four-acid digestion, whilst total
significantly from one area of the resource to another (Jansen and carbonate was determined with the LECO carbon analyzer. A subset of
Taylor, 2003). The use of non-variable parameters for each ore can samples was later sent for X-Ray Diffraction (XRD) using clay separation,
result in either the under- or overestimation of operational costs. at 2 μm, followed by ethylene glycol addition and high temperature heat
Therefore, the ability to model GAC from fundamental ore characteris­ treatment (400 and 550 ◦ C) to differentiate clay species.
tics represents a useful tool since a model can facilitate the review of a
range of possibilities (Youlton and Kinnaird, 2013). A geometallurgical 3. Results and discussion
approach is outlined in this paper, using a pre-operational program of
well-designed test work to develop a comprehensive understanding of Reconciliation of calculated QEMSCAN results to independent
the behaviour of ores, allowing short- and long-term gangue acid con­ chemical analyses for target elements, copper (CuT), iron (FeT), sulfur
sumption to be estimated within a major copper deposit. (ST) and silicon (Si), show good agreement indicating high quality
mineralogy (Fig. 1). Correlation coefficients range from 94 to 96%, apart
2. Methods from iron, at 88%. However, this is understandable in the context of a
much more complex elemental deportment with possible occurrences of
Forty-eight samples were chosen to represent the spatial and iron as copper sulfides (chalcopyrite, bornite), sulfide gangue (pyrite),
mineralogical variability of future leach feed. Ore selection leveraged and iron oxides (goethite, hematite, magnetite, ilmenite), as well as
information available in the resource block model targeting the antici­ various other carbonate (siderite), silicate (illite, biotite, chlorite,
pated range of major acid consuming minerals. Continuous lengths of hornblende, epidote, augite), and sulphate species (jarosite). The iron
half diamond drill core were obtained from various locations, in zones of content of certain minerals varies significantly within the resource,
the deposit above cut-off grade, with the aim of minimizing dilution in further complicating the calculation of elemental abundance from
terms of mineralization, lithology, and alteration type. Tall columns of 6 mineralogy.
m in height were run, requiring large masses of 180 kg, plus an addi­ Carbonate data by LECO also reconcile well with the FeOx/CO3
tional 30 kg for characterization. Feed was prepared by stage-crushing group, from QEMSCAN, showing a strong linear correlation, providing
samples to the necessary size for this process (P80 ¾”). Standard con­ further validation of analytical data. Note: CO32− is abbreviated to CO3
ditions were considered for both agglomeration and irrigation to throughout the paper whilst calcite also occurs locally in the ore body.

2
M. Thomas Minerals Engineering 171 (2021) 107090

Fig. 1. Reconciliation of QEMSCAN vs chemical assays for target elements (left) and total carbonate by LECO vs the FeOx/CO3 group (right).

Goethite, hematite, and magnetite, often present a small carbon peak in X-Ray Diffraction (XRD) is sometimes needed in conjunction with
the EDX (Energy Dispersive X-ray) spectra, due to the requirement for other techniques, like electron microprobe, to distinguish minerals that
sample blocks to be carbon coated, meaning accurate discrimination of are grouped by QEMSCAN for their similar chemical compositions
siderite from iron oxides is challenging during automated analysis. (Ayling et al., 2012). Clay separation, glycol addition, and heat treat­
Nevertheless, limonites such as goethite, will consume acid by break­ ment were used alongside XRD to determine the specific make-up of
down to ferric ions in solution (Jansen and Taylor, 2003) i.e., like “kaolinite/clays”. Kaolinite was found to be the dominant species in this
siderite these minerals can represent a significant source of reagent instance, tending to be of moderate abundance, but occasionally
consumption. Both chemical and mineralogical information regarding reaching more than 20 wt%. No significant levels of pyrophyllite or
carbonates is later used in the prediction of gangue acid consumption to swelling clays were identified. This is beneficial since smectite has the
assess the most reliable source of data. capability to hold acid in its structure (Chetty 2018). Other important
In Fig. 2, quantitative mineralogy is ordered by three types of acid consuming minerals are chlorite, albeit at minor levels, iron-rich
alteration since this demonstrates the main control on the distribution of carbonates (siderite) and some calcite in low abundance, also, 30% of
major acid consumers within the ore body. Potassic biotite (Pot Bio) the samples contain more than 1.5 wt% FeOx/CO3 and, based on the
contains low quartz, with typically 8 to 12 wt% biotite, whilst two other LECO data, CO3 is approximately 3 to 4 times more abundant in the
types of ore generally contain<2 wt% of this mineral. Plagioclase ac­ potassic biotite and feldspar units versus sericite alteration.
counts for 10 to 15 wt% of the potassic feldspar unit (Pot Flds) while Grain size distributions of non-sulfide gangue (NSG), mainly con­
total feldspars normally comprise upwards of 30 wt%. Phyllic ore (Ser) sisting of silicates, are similar in potassic biotite (Pot Bio) and sericite
is characterized by a significant muscovite/sericite component along­ (Ser) units, whereas those of potassic feldspar (Pot Flds) are larger
side relatively high levels of quartz and pyrite. Sericite is not a true (Fig. 3). Greater levels of quartz and feldspars appear to be related to
mineralogical definition but is often used to describe fine-grained white coarser-grained textures. Relatively fine, although variable, textures in
mica, following the secondary alteration of feldspars, which includes combination with higher contents of ferromagnesium minerals and
both muscovite and illite. These minerals are grouped due to composi­ carbonates, suggests that the potassic biotite unit should consume more
tional similarities. acid. This is confirmed by results of the column program (Table 1). The

Fig. 2. QEMSCAN modal mineralogy of 48 crushed ore samples by alteration unit.

3
M. Thomas Minerals Engineering 171 (2021) 107090

Fig. 3. Cumulative distribution frequency (CDF) of pyrite (left); and NSG or non-sulfide gangue (right) by unit.

Table 1
Summary of gangue acid consumption statistics by ore type.
Variable Unit Count Mean SE Mean StDev Min Q1 Median Q3 Max

GAC (kg/t) Pot Bio 15 27.3 0.3 1.1 25.7 26.3 27.4 28.2 29.4
Pot Flds 11 22.9 1.2 4.0 15.5 19.6 24.5 26.0 28.4
Ser 22 19.0 0.7 3.4 13.1 17.2 19.4 21.3 25.5

low variability of GAC in the potassic biotite unit is due to consistently formation. Jarosite precipitation causes acid to be regenerated, or
high abundances of carbonates, biotite, and chlorite. released back into the system, as described by the following reaction,
Carbonate measured by LECO influences the degree of dissolution in where M represents a monovalent cation from the group comprising
columns (see left-hand graph of Fig. 4), as determined by comparison of Na+, NH4+, K+, Ag+, Rb+, 0.5Pb2+ and H3O+ (Monhemius, 2016) but
corresponding pre- and post-leach samples, using 2% mass loss. A ideally K+ or H3O+ in most circumstances (Smith et al., 2006):
dissolution of 60 to 80% is typical for CO3 above 0.3%, whereas < 60%
3Fe3+ + 2SO2−4 + M + + 6H2 O = MFe3 (SO4 )2 (OH)6 + 6H +
is observed at < 0.3% CO3, suggesting a degree of locking. Optical- and
scanning electron microscopy (SEM), alongside a review of QEMSCAN Concentrations of potassium (K), iron, mainly as ferric (Fe3+), and
mineral maps, highlighted an association between siderite and quartz sulphate (SO4) in PLS decrease simultaneously early in the cycle (Fig. 6).
which could explain why a proportion of carbonate does not dissolve. This behavior suggests that significant potassium jarosite KFe3(­
Quartz is impervious and unreactive in leaching meaning that access to SO4)2(OH)6 precipitated. Siderite dissolution will increase iron con­
grains encapsulated within this mineral will only be possible if fractures centration, and local pH, potentially enabling jarosite formation.
are present. The potassic feldspar unit presents markedly higher car­ Short-term GAC was calculated using CO3 head grade, the average
bonate availability when feed grade is over 0.3% vs the other two dissolution factor for each grade/unit, and a consumption factor of 16
alterations. kg/t H2SO4 per 1% CO3. For example, sample 12 contained 1.4% CO3,
Jarosite is typically <0.1 wt% of the feed, however, this increases meaning a dissolution factor of 68% was applied based on alteration
during leaching as a function of carbonate content (Fig. 5). It appears type at a grade over 0.3%, giving rise to a short-term GAC of 15.1 kg/t
that higher carbonate generally results in the production of more jar­ (Eq.1).
osite. This is interpreted as occurring due to rapid initial acid con­
sumption when raffinate is exposed to siderite grains, creating localized Short − termGAC(kg/t) = 16 × (CO3 f eed
areas of high pH, either within coarser particles and/or isolated zones of × (AVG CO3 diss% @0.3%/100)) (Eq.1)
the columns experiencing slow solution transport, resulting in jarosite
Acid regeneration during precipitate formation is determined

Fig. 4. Relationship between CO3 dissolution and CO3 content (left) and boxplot of CO3 dissolution at a threshold of 0.3% CO3 in the feed (right).

4
M. Thomas Minerals Engineering 171 (2021) 107090

Fig. 5. Relationship between jarosite abundance in leached residues vs CO3 in the ore (left) and estimated short term GAC via CO3 vs acid production due to jarosite
precipitation (right).

Fig. 6. Average concentrations of FeT, Fe2+ and Fe3+ (left) and K vs SO4 (right) in column PLS over time.

through a use of a parameter, describing the amount of acid released in inputs (Table 2), taking final GAC values from column tests, minus the
this reaction per unit mineral (-0.587 kg H2SO4/kg K-Fe-sulfate), com­ calculated short-term GAC (Eq.1), adjusted for the level of acid released
bined with an estimate of jarosite precipitation, based upon the rela­ back into the system during jarosite precipitation (Eq.2).
tionship in the left-hand graph of Fig. 5 (Eq.2). Therefore, for
Long − term GAC calc = Final GAC column − Eq.1 + Eq.2 (Eq.3)
approximately 0.7% of CO3 in the feed, 1.0 wt% of jarosite is produced
or 10 kg mineral per ton of ore, subsequently liberating ~ 6 kg of acid Pearson correlation results in Table 3 suggest that the strongest
back into the system. positive drivers of long-term GAC, with r values from ~ 0.5 to 0.8, are
biotite and chlorite, the equivalent chemical variables being magnesium
Acid Regeneration via Jarosite(kg/t) = 5.87(1.06 × CO3 f eed + 0.27)
(Mg) and the ratio of magnesium to magnesium plus iron (Mg/(Mg +
(Eq.2)
Fe)). Fig. 7 shows that the Mg/(Mg + Fe) ratio is closely related to the
Equation (2), when applied to sample 12 (at 1.4% CO3) defines the combined abundance of biotite and chlorite, furthermore, this variable
contribution of acid regenerated during jarosite formation as 5.87 representing a major control of overall GAC alongside CO3.
(1.06*1.4 + 0.27) or 10.3 kg/t. A summary of the previous calculation The best subsets for the prediction of long-term GAC are outlined in
steps is presented in the right-hand graph of Fig. 5. Determination of Table 4. Carbonate data from LECO were chosen to define jarosite pre­
long-term gangue acid consumption (Eq.3) incorporated both previous cipitation, earlier in the study, but the FeOx/CO3 group is utilized here
to firstly assess a model based on quantitative mineralogy alone.

Table 2
Estimates of short-term GAC (St GAC) and acid regeneration from jarosite precipitation (Jar H+ ), by unit, as a function of the CO3 content of the feed, with long-term
GAC (Lt GAC) by difference based on the overall GAC in columns.
Unit GAC (kg/t) Count Mean StDev Min Q1 Median Q3 Max

Pot Bio St GAC 15 11.6 9.1 0.6 0.9 10.5 20.3 28.7
Jar H+ 15 8.4 5.0 2.4 2.8 7.6 13.3 18.0
Lt GAC 15 24.1 3.7 17.1 20.8 24.1 27.5 28.7
Pot Flds St GAC 11 10.3 11.0 0.6 0.8 6.9 13.2 33.9
Jar H+ 11 6.8 5.1 2.4 2.7 5.0 8.0 18.0
Lt GAC 11 19.4 3.9 12.4 17.3 19.7 21.0 26.5
Ser St GAC 22 1.8 2.4 0.5 0.6 0.7 1.4 9.3
Jar H+ 22 3.2 1.3 2.3 2.4 2.6 3.4 7.4
Lt GAC 22 20.4 3.1 14.9 18.8 21.0 21.8 26.5

5
M. Thomas Minerals Engineering 171 (2021) 107090

Table 3
Pearson correlations of long-term GAC with key mineralogical (left) and chemical (right) variables.
Mineralogy Lt FeOx K-feld Plag Ser Kao Bio Chemistry Lt CO3 Al Ca Fe K Mg
GAC /CO3 GAC

FeOx r − 0.22 CO3 r − 0.29


/CO3 p 0.14 p 0.05
K-feld r 0.42 0.19 Al r 0.17 − 0.09
p 0.00 0.19 p 0.25 0.57
Plag r 0.20 0.32 0.58 Ca r 0.23 0.45 − 0.15
p 0.17 0.03 0.00 p 0.12 0.00 0.32
Ser r − 0.25 − 0.47 − 0.69 − 0.82 Fe r − 0.05 0.19 0.42 0.10
p 0.09 0.00 0.00 0.00 p 0.76 0.20 0.00 0.52
Kao r 0.17 0.26 0.01 − 0.17 − 0.25 K r 0.18 − 0.39 0.42 − 0.23 0.25
p 0.26 0.08 0.97 0.26 0.08 p 0.23 0.01 0.00 0.12 0.09
Bio r 0.52 0.47 0.25 0.22 − 0.43 0.33 Mg r 0.58 0.35 0.31 0.59 0.48 0.00
p 0.00 0.00 0.09 0.13 0.00 0.02 p 0.00 0.01 0.03 0.00 0.00 1.00
Chl r 0.68 0.10 0.51 0.47 − 0.45 − 0.15 0.58 Mg/ (Mg r 0.76 0.27 0.12 0.63 0.06 − 0.09 0.88
p 0.00 0.52 0.00 0.00 0.00 0.32 0.00 + Fe) p 0.00 0.07 0.41 0.00 0.70 0.56 0.00

Fig. 7. Scatterplot of Mg/(Mg + Fe) ratio vs the abundance of biotite plus chlorite (left) and overall GAC showing that the Mg(Mg + Fe) ratio has a strong influence
on the process when combined with CO3 (right).

Table 4
Best subsets for long-term GAC modelling using mineralogical predictors.
Model Name Major Statistics Predictors Predictor Validity Residuals Model Valid (Y/N)

R2 R2_adj S FeOx/ CO3 K- Plag Bio Chl p_max (<0.05) VIF_max (<5) RES_p-value (greater than0.05)
feld

1a_min 46 45 2.9 X 0.00 1.00 0.37 Y


1b_min 27 26 3.4 X 0.00 1.00 0.11 Y
1c_min 18 16 3.6 X 0.00 1.00 0.50 Y
2a_min 54 52 2.7 X X 0.00 1.28 0.14 Y
2b_min 54 52 2.7 X X 0.01 1.01 0.57 Y
2c_min 49 46 2.9 X X 0.14 1.49 0.80 N
3a_min 67 64 2.3 X X X 0.00 1.33 0.12 Y
3b_min 66 64 2.3 X X X 0.00 2.02 0.89 Y
3c_min 59 56 2.6 X X X 0.03 1.37 0.16 Y
4a_min 70 68 2.2 X X X X 0.02 2.11 0.91 Y
4b_min 67 64 2.4 X X X X 0.69 1.61 0.12 N
4c_min 67 63 2.4 X X X X 0.57 2.18 0.93 N

Intermediate plagioclase species (Plag), mainly andesine, as well as k-


feldspar (K-feld) are typically considered to be moderate long-term acid LtGAC pred min = 17.63 − 1.80FeOx/CO3 + 0.16Kf eld + 0.46Bio + 1.94Chl
consuming minerals. These were selected in conjunction with biotite (Eq.4)
(Bio) and chlorite (Chl) which are usually more significant sources of Reconciliation of real vs predicted long-term GAC results via model
GAC in heap leaching (Chetty, 2018). Other relevant work has reported 4a_min, presented in Fig. 8, shows a moderate to strong correlation and
that biotite may totally dissolve, whilst chlorite is also highly soluble, normally distributed residuals. The negative coefficient for the carbon­
almost completely dissolving in strong acidic environments (Snäll and ate group in Eq.4 indicates that lower levels of long-term acid con­
Liljefors, 2000). Pyrite is not relevant here as the Eh-pH conditions sumption tend to occur when more CO3 is present. This could be due to a
associated with this process mean that it remains generally inert. In significant proportion of the acid reacting during the initial stages of
terms of long-term GAC predictability, using mineralogy, model 4a_min leach, especially curing, which appears to limit the later impact of sili­
offers the highest adjusted correlation coefficient (R2_adj) alongside the cates. It is known that the degree of acid consumption associated with
lowest standard error of regression (S):

6
M. Thomas Minerals Engineering 171 (2021) 107090

Fig. 8. Estimated vs predicted long term GAC using model 4a_min (left) and corresponding residual plots (right).

silicate minerals increases when acid concentration is high (Van Staden Carbonate dissolution values (CO3_diss%) are 29% when CO3 <
et al., 2009), so on the contrary, lower acid availability could result in 0.3%, and 68% (Pot Bio), 80% (Pot Flds) or 63% (Ser) when
silicate gangue exhibiting a less aggressive behavior. The relative effect CO3>=0.3%, with Mg, Fe and CO3 measured in percent.
of this mechanism may be accentuated since a significant proportion of
Overall GAC(kg/t)= (16(CO3 × (CO3 diss%/100)) − (5.87(1.06 × CO3
the acid was added during agglomeration.
Data mining efforts were now focussed on permitting long-term GAC + 0.27)) + (14.83 − 2.88CO3 + 58.27Mg/(Mg + Fe))
to be predicted from chemistry, instead of mineralogy, as this type of
Carbonates are responsible for just over a third of overall GAC in the
analysis is much cheaper, and more feasible, if the intent is to apply it to
sample set (Fig. 11). Potassic biotite alteration has the most significant
the whole 3D block model. Table 5 shows that options using 3 or 4
short- and long-term gangue acid consumption, whilst the lowest
variables have R2_adj values over 80%, however, all fail validation due
average long-term GAC occurs in the potassic feldspar unit, and the
to the p-values of corresponding inputs exceeding 0.05. Variance Infla­
lowest short-term GAC in the phyllic ore. However, overall consumption
tion Factors (VIF) are often over the target value of 5, meaning that
is lower than initially anticipated due to the impact of jarosite precipi­
predictors are highly intercorrelated, thereby resulting in poor quality
tation; in fact, estimates of long-term GAC would be sporadically
estimates of regression coefficients. Model 2a_ch is the best option
negative in some samples if no attempts were made to correct for this
passing validation where R2_adj is 82% and S at 1.6. CO3 constitutes
mechanism. Acid regeneration must therefore be contemplated to make
mainly siderite in this deposit whereas the ratio Mg/(Mg + Fe) is
a sensible evaluation of this system. A summary of final model perfor­
indicative of ferromagnesium mineral content. The statistics of the
mance (Eq.6) shows excellent agreement with real data where averages
chemistry based long-term GAC model (Fig. 9) significantly improve on
for each unit fall within ± 1% of column results (Table 6).
those of the mineralogy case (4a_min) with 2a_ch being defined as:
An independent test set, comprising approximately an additional
LtGACpredchem = 14.83 − 2.88CO3 + 58.27Mg/(Mg + Fe) (Eq.5) 15% of samples, was used to check the predictability of the model.
Fig. 12 shows that the statistics of the test set are closely aligned with
Determination of overall GAC combines the previous equations to
those of the modelling data. Validation against heaps is not yet possible,
incorporate the effects of short- and long-term components while dis­
as the project is still in the developmental stage, meaning that the in­
counting the acid produced by jarosite formation. Only chemical pre­
dustrial process has not been constructed at this point.
dictors are considered from now on. Essentially, ICP analysis for Fe and
Following the completion of proof-of-concept, the drill core database
Mg, plus CO3 derived by LECO, prove sufficient to confidently model
was populated to increase the density of information, using bench-scale
performance trends. Equation (6) (Fig. 10) provides elatively precise
composites collected on a regular 75x75m grid with iron, magnesium
estimations, as the value of S is small (just over 1.5), whilst a large
and carbonate being measured by ICP and LECO techniques. A series of
proportion of the variance in overall GAC is explained by the model (R2
geostatistical methods were reviewed to interpolate key inputs,
88%):
including ordinary- and co-Kriging, which then allowed estimations of
Overall GAC(kg/t) = Eq.1 − Eq.2 + Eq.5 (Eq.6) gangue acid consumption to be made via a script in Vulcan, linking

Table 5
Best subsets for long-term GAC modelling using chemical predictors.
Model Name Major Statistics Predictors Predictor Validity Residuals Model Valid (Y/N)

R2 R2_adj S CO3 Fe K Mg Mg/ (Mg + Fe) p_max (<0.05) VIF_max (<5) RES_p-value (greater than0.05)

1a_ch 57 56 2.6 X 0.00 1.00 0.40 Y


1b_ch 33 32 3.2 X 0.00 1.00 0.46 Y
1c_ch 8 6 3.8 X 0.00 1.00 0.33 Y
2a_ch 83 82 1.6 X X 0.00 1.08 0.39 Y
2b_ch 63 61 2.4 X X 0.01 1.01 0.94 Y
2c_ch 61 59 2.5 X X 0.00 1.14 0.29 Y
3a_ch 83 82 1.7 X X X 0.38 1.26 0.30 N
3b_ch 83 82 1.7 X X X 0.40 4.67 0.88 N
3c_ch 83 82 1.7 X X X 0.97 1.08 0.38 N
4a_ch 85 83 1.6 X X X X 0.06 25.49 0.18 N
4b_ch 84 82 1.6 X X X X 0.25 5.08 0.51 N
4c_ch 83 82 1.7 X X X X 0.77 1.39 0.41 N

7
M. Thomas Minerals Engineering 171 (2021) 107090

Fig. 9. Estimated vs predicted long term GAC using model 2a_ch (left) and corresponding residual plots (right).

Fig. 10. Real vs predicted overall GAC (left) and model residual plots (right).

Fig. 11. Modelled short- (St GAC) and long-term (Lt GAC) gangue acid consumption adjusted for jarosite (Jar H+ ) precipitation (left) and comparison of averages for
predicted vs measured GAC in columns (right) by unit.

predictive equations to the block model. Large-scale metallurgical


Table 6
piloting provided additional support to define a scaling factor (0.8) for
Summary of staged-gangue acid consumption using chemical predictors
the conversion of column performance to that of industrial heaps.
(GAC_model), including a correction for jarosite precipitation, compared to
average column performance (GAC_cols) with values in kg/t. Geometallurgical predictors were capped in certain instances assuring
alignment with the training dataset. <1% of blocks in the resource
Eq.1 Eq.2 Eq.5 Eq.6 (1–2 + 5)
needed the carbonate content to be restricted at an upper threshold of
Alteration Unit St_GAC Jar_H+ Lt_GAC GAC_model GAC_cols 2.5%, whilst estimations of long-term gangue acid consumption were
Potassic Biotite 11.6 8.4 24.2 27.4 27.3
limited to the scaled maximum and minimum values from the column
Potassic Feldspar 10.3 6.8 19.5 23.0 22.9
Sericite 1.8 3.2 20.3 18.9 19.0 results, affecting<5% of blocks. Fig. 13 shows the distribution of short-
term, long-term, and overall GAC, in the middle part of the deposit, as
well as the amount of acid released due to jarosite precipitation. Clear

8
M. Thomas Minerals Engineering 171 (2021) 107090

Fig. 12. Real vs predicted GAC (left) and model residual plots (right) for an independent validation set.

Fig. 13. Plan view of the central area of the deposit, using the 3D block model at a cut-off 0.2% CuT, showing the distribution: of short-term gangue acid con­
sumption, St GAC (left); acid regeneration via jarosite precipitation, Jar H+ (left-center); long-term gangue acid consumption, Lt GAC (right-center); and overall
gangue acid consumption, GAC (right).

zonation can be observed, in terms of higher short-term GAC, mainly in to optimize acid distribution and leaching. Whilst this phase of model­
the upper and lower central sections of Fig. 13, whilst a significant area ling described the expected performance, under standard conditions,
presenting greater levels of long-term GAC mainly occurs to the center optimization opportunities were highlighted, in more challenging areas
and upper right quadrants, and, in the far bottom left to a lesser extent. of the deposit, that required further investigation.
This study clearly shows how a holistic knowledge of the ore body
4. Conclusions may provide an in-depth understanding of future processing response
prior to key investment decisions relating to construction and project
The amount of carbonate available for dissolution in leaching was implementation. Many expensive mistakes have been made in the in­
found to be related to a combination of mineral feed grade and alteration dustry often leading to early mine closure and significant financial los­
type in this case. Once an understanding of liberation was established, ses. The hope of this paper is to outline a methodology capable of
dissolution factors could be applied, together with carbonate feed con­ mitigating this risk somewhat.
tent, to calculate short-term gangue acid consumption based on stoi­
chiometry. Jarosite precipitation, and acid regeneration (Eq.2), was also CRediT authorship contribution statement
correlated to carbonate abundance. Both estimates, when linked to
overall column results, describe the long-term gangue acid consumption Mark Thomas: Conceptualization, Data curation, Formal analysis,
component by difference. Biotite (Bio), chlorite (Chl), orthoclase (K- Investigation, Methodology, Writing – original draft, Writing – review &
feld) and FeOx/CO3 are the major mineralogical drivers to predict long- editing.
term GAC, supporting the development of an alternative option using
equivalent chemical variables to improve on model quality. Through a Declaration of Competing Interest
process of exploratory data analysis, inputs were simplified to three
chemical predictors (CO3, Fe and Mg), providing a robust foundation for The authors declare that they have no known competing financial
the estimation of GAC. interests or personal relationships that could have appeared to influence
A systematic drilling campaign and subsequent analysis of samples the work reported in this paper.
by LECO and ICP, followed, using appropriate geostatistical techniques,
permitting key variables to be spatially distributed. Each relevant block Acknowledgements
then contained the necessary inputs to define GAC when linked to the
geometallurgical equations outlined in this paper. The three-stage Great appreciation is given to all colleagues, clients and mentors who
approach enables agglomeration and curing to be carried out correctly have provided invaluable support and guidance, over many years,
in order to obtain anticipated leaching performance. Knowledge of enabling the knowledge required to write this paper to be assimilated.
gangue acid consumption variability also creates a solid basis for the Individuals contributing to the review process of this manuscript and
assessment of future operational costs over time. It is important to providing excellent supporting information, as well as valuable insights,
comprehensively understand the relationship between the short- and include Dr Dave Dew and Prof Jens Andersen from the Camborne School
long-term acid demand of the ore, in advance of mining and processing, of Mines.

9
M. Thomas Minerals Engineering 171 (2021) 107090

References Montealegre, R., 2002. Cuprochlor Heap Leaching Process. Minerals Eng. Int. Online -
Hydrometallurgy (accessed 12/01/2021).
Morera-Chavarria, A., Griffioen, J., Behrends, T., 2016. Optimized sequential extraction
Acker, J.G., Bricker, O.P., 1992. The influence of pH on biotite dissolution and alteration
of carbonates: quantification and d13C analysis of calcite, dolomite, and siderite.
kinetics at low temperature. Geoehimica ef Cosmochimica Acia 56, 3073–3092.
Chem. Geol. 443, 146–157.
Ayling, B., Rose, P., Petty, S., Zemach, E., Drakos, P., 2012. QEMSCAN (Quantitative
Nowaczky, K., Domka, F., 2000. Oxidation of pyrite and marcasite by Thiobacillus
Evaluation of Minerals by Scanning Electron Microscope): capability and application
ferrooxidans bacteria. Polish J. Environ. Studies 9 (2), 87–90.
to fracture characterization in geothermal systems. Proceedings of the 37th
Petersen, J., 2016. Heap leaching as a key technology for recovery of values from low-
Workshop on Geothermal Reservoir Engineering. Stanford University (accessed 20/
grade ores – a brief overview. Hydrometallurgy 165 (1), 206–212.
03/2021).
Salgado, C., Perez, C., Alvayai, C., and Zarate, G. 2009. Acid management in heap
Chetty, D., 2018. Acid-gangue interactions in heap leach operations: a review of the role
leaching: are we doing it right? Hydrocopper Conference, Chile, 47-56.
of mineralogy for predicting ore behavior. Minerals 8 (47), 1–11.
Schlesinger, M.E., King, M.J., Sole, K.C., Davenport, W.G., 2011. Extractive Metallurgy of
Free, M.L., 2010. Understanding acid consumption and its relationship with copper
Copper, fifth ed. Elsevier, Amsterdam.
recovery. SME Annual Meeting, Phoenix, AZ, pp. 1–4.
Sinclair, W.D. 2007. Porphyry Deposits. In: Mineral Deposits of Canada - A Synthesis of
Ghorbani, Y., Franzidis, J.P., and Petersen, J. 2015. Heap leaching technology – current
Major Deposit Types, District Metallogeny, the Evolution of Geological Provinces,
state, innovations, and future directions: a review. https://core.ac.uk/download/
and Exploration Methods, 223-243.
pdf/43096092.pdf (accessed 17/01/2021).
Smith, A.M.L., Hudson-Edwards, K.A., Dubbin, W.E., Wright, K., 2006. Dissolution of
Holland, H.D., and Turekian, K.K. 2014. Dissolution of silicate minerals. In: Treatise on
jarosite [KFe3(SO4)2(OH)6] at pH 2 and 8: insights from batch experiments and
Geochemistry, 1-16, 2nd Ed., ISBN: 9780080983004.
computational modelling. Geochimica Et Cosmochim. Acta 70 (3), 608–621.
Jansen, M., and Taylor, A. 2003. Overview of gangue mineralogy issues in oxide copper
Snäll, S., Liljefors, T., 2000. Leachability of major elements from minerals in strong acids.
heap leaching. https://www.911metallurgist.com/blog/wp-content/uploads/2016/
J. Geochem. Exploration 71 (1), 1–12.
06/HEAP-LEACHING-Mineralogy.pdf (accessed 16/03/2021).
Sutherland, D., 2007. Estimation of mineral grain size using quantitative mineralogy.
John, L.W., 2011. The art of heap leaching – the fundamentals. In: The Southern African
Min. Eng. 20, 452–460.
Institute of Mining and Metallurgy, percolation leaching: the status globally and in
Tan, H., Skinner, W., Addai-Mensah, J., 2012. Leaching behaviour of low and high Fe-
South Africa, pp. 17–42.
substituted chlorite clay minerals at low pH. Hydrometallurgy 125 (6), 100–108.
Jones, M.P., 1987. Applied mineralogy: a quantitative approach. Graham and Trotman,
Van Staden, P.J., Robertson, S.W., Gericke, M., Neale, J.W., Seyedbagheri, A., 2009.
London, pp. 70–80.
Maximising the value derived from laboratory test work towards heap leaching
Lu, J., Dreisinger, D., West-Sells, P., 2017. Acid curing and agglomeration for heap
design. In: The Southern African Institute of Mining and Metallurgy Base Metals
leaching. Hydrometallurgy 167, 30–35.
Conference, pp. 185–200.
Monhemius, J. 2016. The Iron Elephant - A brief history of hydrometallurgists’ struggles
Youlton, B.J., Kinnaird, J.A., 2013. Gangue-reagent interactions during acid leaching of
with Element No.26. XXVIII International Mineral Processing Congress, Quebec,
uranium. Min. Eng. 52, 62–73.
Canada, 1-14.

10

You might also like