minerals

Article
Modeling the Effect of Composition and
Temperature on the Conductivity of Synthetic
Copper Electrorefining Electrolyte
Taina Kalliomäki *, Jari Aromaa and Mari Lundström
Department of Materials Science and Engineering, School of Chemical Technology, Aalto University,
P.O. Box 16200, Aalto FI-00076, Finland; jari.aromaa@aalto.fi (J.A.); mari.lundstrom@aalto.fi (M.L.)
* Correspondence: Taina.Kalliomaki@aalto.fi; Tel.: +358-40-734-8658

Academic Editor: William Skinner

Received: 1 March 2016; Accepted: 13 June 2016; Published: 24 June 2016

Abstract: The physico-chemical properties of the copper electrolyte significantly affect the energy
consumption of the electrorefining process and the quality of the cathode product. Favorable
conditions for electrorefining processes are typically achieved by keeping both the electrolyte
conductivity and diffusion coefficient of Cu(II) high, while ensuring low electrolyte viscosity. In
this work the conductivity of the copper electrorefining electrolyte was investigated as a function of
temperature (50–70 ˝ C) and concentrations of copper (Cu(II), 40–60 g/L), nickel (Ni(II), 0–20 g/L),
arsenic (As(III), 0–30 g/L) and sulfuric acid (160–220 g/L). In total 165 different combinations of these
factors were studied. The results were treated using factorial analysis, and as a result, four electrolyte
conductivity models were built up. Models were constructed both with and without arsenic as the
presence of As(III) appeared to cause non-linearity in some factor effects and thus impacted the
conductivity in more complex ways than previously detailed in literature. In all models the combined
effect of factors was shown to be minor when compared to the effect of single factors. Conductivity
was shown to increase when copper, nickel and arsenic concentrations were decreased and increase
with increased temperature and acidity. Moreover, the arsenic concentration was shown to decrease
the level of conductivity more than previously suggested in the literature.

Keywords: copper electrorefining electrolyte; conductivity; conductivity model

1. Introduction
Copper electrorefining is the most common method for producing high-purity copper [1]. The
first refinery (Pembrey Copper Works) was established in 1869 following the first patent for commercial
electrorefining developed by James Elkington in 1865 [2,3]. Since then, the process has been developed
further as the result of both research and improved industrial practices [2]. In the copper electrorefining
process, copper is dissolved from impure copper anodes into the electrolyte bath and then subsequently
deposited on to cathodes as high-purity copper [4].
Industrial copper refining electrolytes mainly consist of water, copper sulfate, sulfuric acid, with
additional leveling/grain-refining agents—to obtain smoother and denser cathode deposits—and
process impurities [4,5]. These impurities commonly consist of nickel, arsenic and iron, along with
smaller amounts of bismuth, antimony and chloride that dissolve into the electrolyte from the anode.
The physico-chemical properties of the copper electrolyte significantly affect the yield of cathodic
copper in the electrorefining process and include four main physico-chemical properties: conductivity,
density, viscosity and the diffusion coefficient of the cupric ion (Cu(II)) [5–11]. The best yield of
copper can be obtained by keeping the electrolyte viscosity low [5] while ensuring a high diffusion
coefficient [7] and electrical conductivity [5]. These properties of the copper electrolyte are strongly

Minerals 2016, 6, 59; doi:10.3390/min6030059 www.mdpi.com/journal/minerals

Minerals 2016, 6, 59 2 of 11

influenced by composition and temperature: Increased temperature, for example, lowers electrolyte
density and viscosity [6] while enhancing the rate of chemical reactions [4]. On the other hand, a
too-high temperature results in unnecessary energy costs and excessive electrolyte bath evaporation [4].
In contrast, an electrolyte composition with a high concentration of Cu(II), Ni(II) and sulfuric acid
makes the electrolyte denser and more viscous [6,12]. An increase in viscosity decreases the diffusion
coefficient of Cu(II) but a high concentration of Cu(II) also increases the limiting current density [7],
giving the possibility for a higher deposition rate. Moreover, an increase in the concentration of sulfuric
acid leads to enhanced conductivity [5–7]. As a result, it is important to thoroughly determine the
effects of these parameters in order to optimize the yield of cathode copper.
Conductivity is an important solution property [13] as it affects the electrical energy consumption
of the electrorefining process [5,6]. Therefore, when optimizing the refining process by altering the
temperature or composition, conductivity is a good value to be controlled either by measuring it or
defining it with an applicable model.
The aim of this study is to construct accurate mathematical models for the conductivity, taking
into account the effect of temperature as well as solution sulfuric acid and metal concentrations.
Furthermore, a particular focus is also paid to the effect of typical impurities such as nickel and arsenic,
originating from increasingly impure raw materials in addition to copper. The design of experiments
is carried out with full factorial design by MODDE software (MKS Data Analytics Solutions, Malmö,
Sweden) in order to build up a model that reflects copper electrorefining conductivity as a function of
all the previously mentioned parameters.

2. Materials and Methods

Electrolytes were prepared from copper sulfate (CuSO4 ¨ 5H2 O, min. 98%, VWR Chemicals,
Radnor, PA, USA), nickel sulfate (NiSO4 ¨ 6H2 O, min. 98%, VWR Chemicals), sulfuric acid (H2 SO4 ,
95%–98%, Sigma-Aldrich, St. Louis, MO, USA), arsenic acid solution (from Boliden Harjavalta
Copper electrorefinery, Harjavalta, Finland, containing [As] = 151,700 mg/L, [Cu] = 4794 mg/L,
[Sb] = 3954 mg/L, [Ni] = 1688 mg/L, [Bi] = 6.2 mg/L, [Te] = 18.6 mg/L, [Pb] = 29 mg/L,
[Ag] = 0.16 mg/L) and distilled water. Cu and Ni contents in arsenic acid were also taken into
account when preparing the electrolytes. Table 1 summarizes the experimental factors and solution
parameters studied, 33 different solution parameter combinations were investigated at five different
temperatures and in total 165 different combinations were studied. No extra additives such as glue,
thiourea or chloride ions were used.

Table 1. Parameters investigated affecting copper electrorefining electrolyte—factors ([Cu(II)], [H2 SO4 ],
[Ni(II)], [As(III)] and temperature, T) and their levels. Arsenic was adjusted by arsenic acid (Cu(II) and
Ni(II) in acid taken into account).

Factor Levels
Cu(II) 40; 50; 60 g/L
H2 SO4 160; 180; 200; 220 g/L
Ni(II) 0; 10; 20 g/L
As(III) 0; 15; 30 g/L
T 50; 55; 60; 65; 70 ˝C

Conductivity measurements were carried out using a Knick Portamess® 913 Cond conductivity
meter produced by Knick Elektronische Messgeräte GmbH & Co. KG (Berlin, Germany). The meter
was used with a four-electrode sensor (ZU 6985) which has glass/platinum measuring system and
glass casing tube. All electrolytes were heated prior to measurement in a jacketed cell using a MGW
Lauda MT/M3 circulating water bath (Figure 1) (LAUDA, Lauda-Königshofen, Germany). During
the heating or between the measurements, the cell was completely covered and all the holes in the
Minerals 2016, 6, 59 3 of 11
Minerals 2016, 6, 59 3 of 10

(DOE)
lid wereand multivariate
plugged data evaporation
to prevent analysis. Theandexperiments werewater
consequently designed
loss.by defining
The factors,
electrolyte wasresponses
stirred at
and levels
400 rpm of the
using factors using
a magnetic the full factorial design.
stirrer.

1. Schematic
Figure 1.
Figure Schematic ofof the
the experimental set-up used
experimental set-up used for
for conductivity measurements: (a)
conductivity measurements: (a) water
water bath;
bath;
(b)
(b) magnetic
magnetic stirrer;
stirrer; (c)
(c) jacketed
jacketed cell;
cell; and
and (d)
(d) conductivity
conductivity sensor.
sensor.

Prior
Data to modeling,
analysis and the raw data design
experiment was explored and evaluated
were carried out using withthescatter plots and histograms.
Modeling design tool
The
MODDE models were constructed
8 software (MKS Data according to the data
Analytics Solutions, and evaluated
Malmö, Sweden) forusing regression
design analysis(DOE)
of experiments tools
summary of fit, analysis of variance (ANOVA) and normal probability
and multivariate data analysis. The experiments were designed by defining factors, responses and plot of residuals. In addition,
the designs
levels of the were
factors checked
using theto ensure a low design.
full factorial enough condition number, i.e., the ratio of the minimum
and maximum singular values of the
Prior to modeling, the raw data was explored factors. For a good design thiswith
and evaluated value is lessplots
scatter thanand3, whereas in a
histograms.
bad design
The modelsitwere is over 6 [14]. The according
constructed effect of thetochanges
the datainand
the models
evaluated after refining
using was also
regression determined
analysis tools
by comparison
summary of fit,of the change
analysis in value(ANOVA)
of variance of the condition number.
and normal probability plot of residuals. In addition,
In summary,
the designs were checkedthe parameters
to ensure athatlow describe the model
enough condition are goodness
number, of fit
i.e., the ratio of (R
the2),minimum
goodnessand of
prediction (Q ), model
maximum singular 2 validity
values of theand reproducibility.
factors. Model this
For a good design validity
valueisisbased on the
less than lack of fitinwhich
3, whereas a bad
is a statistical
design it is over F-test where
6 [14]. Thethe model
effect error
of the is compared
changes to replicate
in the models after error
refining[14].was
In aalso
valid model there
determined by
is no lack of fit, and consequently the
comparison of the change in value of the condition number.model validity is high. The reproducibility describes the
variabilities
In summary,in the replicates [14]. The
the parameters thatmodels werethe
describe refined
model to are
maximize
goodness fit (R2 ), goodness
theseofcorrelation coefficients
of
as well as 2
minimize the difference between the R 2 and Q2 values [14]. In a good model the Q2, the
prediction (Q ), model validity and reproducibility. Model validity is based on the lack of fit which is a
model validity
statistical F-testand where thethe
reproducibility
model error is are larger than
compared 0.5, 0.25 and
to replicate error0.5, respectively
[14]. [14]. Inthere
In a valid model addition,
is no
the
lackdifference
of fit, and between
consequentlythe Rthe
2 and Q2 in a good model is less than 0.2–0.3 [14].
model validity is high. The reproducibility describes the variabilities
in the replicates [14]. The models were refined to maximize these correlation coefficients as well as
3. Resultsthe
minimize anddifference
Discussion between the R2 and Q2 values [14]. In a good model the Q2 , the model validity
and the
Thereproducibility
histogram of the aremeasured
larger than 0.5, 0.25 anddata
conductivity 0.5, is
respectively [14]. In2 addition,
shown in Figure and it canthe bedifference
seen that
between the R 2 and Q2 in a good model is less than 0.2–0.3 [14].
the histogram of the conductivity data is slightly skewed. The model validity and efficiency of data
analysis are better when the data in the histogram plot is less skewed. Scatter plots of the raw data
3. Results and Discussion
(Figure 3) showed both linearity and non-linearity which suggests that the relationship between some
The
factors andhistogram of themight
the response measured conductivity
be curved, as welldata is shown
as the in Figure
possibility 2 andmay
that there it can
be be seen that
interactions
the histogram
between of theNevertheless,
the factors. conductivity data
theseisscatter
slightly skewed.
plots Thean
only give model validity and
approximate efficiency
estimation of data
of how the
analysiscan
factors areinfluence
better when the data in the histogram plot is less skewed. Scatter plots of the raw data
the conductivity.
(Figure 3) showed both linearity and non-linearity which suggests that the relationship between some
factors and the response25might be curved, as well as the possibility that there may be interactions
between the factors. Nevertheless,
20 these scatter plots only give an approximate estimation of how the
Count

factors can influence the conductivity.

15
10
5
0
435 465 495 525 555 585 615 645 675 705 735 765 795 825
mS/cm

Figure 2. Histogram of conductivity values of synthetic copper electrorefining electrolytes (for Model 1).
the histogram of the conductivity data is slightly skewed. The model validity and efficiency of data
analysis are better when the data in the histogram plot is less skewed. Scatter plots of the raw data
(Figure 3) showed both linearity and non-linearity which suggests that the relationship between some
factors and the response might be curved, as well as the possibility that there may be interactions
between
Minerals the6,factors.
2016, 59 Nevertheless, these scatter plots only give an approximate estimation of how
4 ofthe
11
factors can influence the conductivity.

25
20

Count
15
10
5
0
435 465 495 525 555 585 615 645 675 705 735 765 795 825
mS/cm

Figure
Figure
Minerals Histogram
2.6,Histogram
2016,2. 59 ofconductivity
of conductivityvalues
valuesof
ofsynthetic
syntheticcopper
copperelectrorefining
electrorefiningelectrolytes
electrolytes(for
(forModel
Model1).1).4 of 10

(a)

(b) (c)
Figure 3.
Figure 3. Scatter
Scatter plots
plots of
of the
the raw
raw data
data for
for synthetic
synthetic copper
copper electrolyte
electrolyte with
with factors
factors (a) T and
(a) T and Ni(II)
Ni(II)
concentration; (b) H 2SO4 and As(III) concentration; and (c) H2SO4 and Cu(II) concentration.
concentration; (b) H2 SO4 and As(III) concentration; and (c) H2 SO4 and Cu(II) concentration.

3.1. Conductivity
3.1. Conductivity Model
Model 1—Untreated
1—Untreated DataData with
with Terms
Termsofof[H
[H2SOSO4]]22 and T22
2 4 and T
In order
In order toto build
build upup the
the first
first model
model that
that describes
describes copper
copper electrolyte
electrolyte conductivity,
conductivity, unscaled
unscaled
conductivitydata
conductivity datawas
wasused.
used.
TheThe model
model waswas constructed
constructed from from
the raw thedata
rawdespite
data despite the skewed
the slightly slightly
skewed nature of the data
nature of the data (Figure 2). (Figure 2).
Coefficients for
Coefficients for unrefined
unrefined synthetic
synthetic copper
copper electrolyte
electrolyte conductivity
conductivity (Model
(Model 1) 1) are
are shown
shown inin
Figure 4.
Figure 4. It
It can be seen
can be seen that
that in
in addition
addition toto the
the factors
factors studied,
studied, thethe combined
combined effect
effect of
of two
two factors
factors
(product) is also investigated. Terms with a high p-value (probability value), in which
(product) is also investigated. Terms with a high p-value (probability value), in which the error bars the error bars
extend over the zero line, need to be excluded from further analysis, e.g., the terms Cu(II)·H
extend over the zero line, need to be excluded from further analysis, e.g., the terms Cu(II)¨ H2 SO4 and 2SO4 and

Cu(II)·T for Model 1. The summary of fit for Model 1 showed that the R2, Q2 and reproducibility were
at a good level, but the model validity was found to be poor.
Model validity was improved by adding the squares of the terms to the model. Using squares
[H2SO4]2 and T2, the irregularities of the normal probability plot could be reduced, although this was
seen not to improve the model validity. Two result series (with a maximum amount of H2SO4, Cu(II)
and Ni(II) as well as a maximum and medium amount of As(III)) were removed from the model, as
it was suspected that these electrolytes were supersaturated and thus responsible for the skewness in
Minerals 2016, 6, 59 5 of 11

Cu(II)¨ T for Model 1. The summary of fit for Model 1 showed that the R2 , Q2 and reproducibility were
at a good
Minerals level,
2016, 6, 59 but the model validity was found to be poor. 5 of 10

Figure 4. Coefficients for unrefined synthetic copper electrolyte conductivity in Model 1. Adj. =
Figure 4. Coefficients for unrefined synthetic copper electrolyte conductivity in Model 1.
adjusted,
Adj. Conf. lev.
= adjusted, = confidence
Conf. lev. = level, DF = degrees
confidence level, ofDF
freedom, RSD =ofresidual
= degrees standard
freedom, RSD deviation.
= residual
standard deviation.
Table 2. R2, Q2, model validity and reproducibility values of the models.

Model validity was improved Model by adding 1 the squares

2 3 terms to
of the 4 the model. Using squares
2 2 R 2 0.9978 0.9985 0.9972 0.9984
[H2 SO4 ] and T , the irregularities of the normal probability plot could be reduced, although this was
seen not to improve the model Q 0.9876 0.9959 (with0.9661 0.9739 amount of H SO , Cu(II)
2
validity. Two result series a maximum 2 4
Model validity 0.3016 0.4598 0.2609 0.4194
and Ni(II) as well as a maximum and medium amount of As(III)) were removed from the model, as it
Reproducibility 0.9997 0.9996 0.9996 0.9996
was suspected that these electrolytes were supersaturated and thus responsible for the skewness in
the results.
Equation (1) for conductivity Model 1 was compiled according to unscaled coefficients.
These modifications resulted in a reasonable model validity as well as slightly higher R2 and Q2 ;
κ = 97.72number
however, the condition − 3.581 increased
[Cu(II)] + from
0.47362.427
[Ni(II)]
(with+ 0.596 [As(III)] +data)
the unrefined 2.945to[H 2SO4] +
4.517—slightly high
0.02396 [Cu(II)][Ni(II)] + 0.006713 2
terms [H[Cu(II)][As(III)] 2
but below 6—as a result of adding the 2 SO4 ] and T . +The 0.01219
model [Ni(II)][As(III)]
had no lack of − fit in this
(1)
phase and 0.02297 [H2SO4][Ni(II)]
the summary − 0.02166
of fit indicated [Hthe
that 2SOmodel
4][As(III)] − 0.01899
is valid (Table[Ni(II)]T
2). + 0.01768 T[As(III)]
+ 0.02754 [H2SO4]T + 2.743 T − 0.005364 [H2SO4]2 − 0.02946 T2
Table 2. R2 , Q2 , model validity and reproducibility values of the models.
where the concentrations are in g/dm3, T is in °C and κ is in mS/cm.
The equation has many terms, and1 all the effects
Model 2 of the factors
3 are not fully4seen in the equation
coefficients. Nonetheless, factors with low p-values indicate that at least H2SO4 seems to have a
R2 0.9978 0.9985 0.9972 0.9984
combined effect with QNi(II),
2 temperature
0.9876 and As(III).
0.9959Data used
0.9661 Model 0.9739
for 1 indicated that Cu(II),
Ni(II) and As(III)Model
lowervalidity
the conductivity
0.3016of the electrolyte
0.4598 while temperature0.4194
0.2609 and H2SO4 increase it,
which is in line with the literature [5,6].
Reproducibility 0.9997 0.9996 0.9996 0.9996

3.2. Conductivity
Equation (1)Model 2—Logarithmic
for conductivity Data
Model with compiled
1 was Terms of [H2SO4]2 and T2
according to unscaled coefficients.
In order to avoid the skewness in Model 1 with untreated data, the second model was
constructed “ 97.72
κ using ´ 3.581 rCupIIqs
logarithmic values` of 0.4736 ` 0.596
rNipIIqsand
conductivity 2.945 rH
rAspIIIqsin`a more
this resulted 2 SO4 s `distributed
normally
histogram.0.02396
TermsrCupIIqsrNipIIqs
with a high p-value
` 0.006713 were removed along
rCupIIqsrAspIIIqs with the
` 0.01219 two result series
rNipIIqsrAspIIIqs ´ of the
(1)
presumably supersaturated electrolytes determined from Model 1. As before, the term squares T2 and
0.02297
[H2SO rH SO4 srNipIIqs
4]2 were 2added
´ 0.02166
to reduce rH2 SO4 srAspIIIqs
the irregularities of the ´ 0.01899
normal rNipIIqsT `
probability 0.01768 TrAspIIIqs `
plot.
Equation (2) for0.02754
the logarithm of the
rH2 SO4 sT conductivity
` 2.743 was rH
T ´ 0.005364 compiled 2 using unscaled
2 coefficients. The
2 SO4 s ´ 0.02946 T
strongest combined effects based on low p-values were shown to be with H2SO4·T, H2SO4·As(III) and
where thehowever,
concentrations are in g/dm 3 , T is in ˝ C and κ is in mS/cm.
T·As(III); as with Model 1, the combined effects were minor compared to the single effects
The equation has many terms, and all the effects of the factors are not fully seen in the equation
of factors.
coefficients. Nonetheless, factors with low p-values indicate that at least H2 SO4 seems to have a
log10(κ) = 2.17388 − 0.0023479 [Cu(II)] − 0.0027733 [Ni(II)] − 0.00073729 [As(III)] +
0.0037764 [H2SO4] − 1.0649 × 10−5 [H2SO4][As(III)] + 2.1627 × 10−5 T[As(III)] + (2)
8.8019 × 10−6 [H2SO4]T + 0.0051846 T − 6.9222 × 10−6 [H2SO4]2 − 3.2506 × 10−5 T2,
Model 2 was regarded as reasonable and better with respect to Model 1, due to the improved
model validity and Q2 value, and Q2 was almost equal to R2 (Table 2). However, the condition number
Minerals 2016, 6, 59 6 of 11

combined effect with Ni(II), temperature and As(III). Data used for Model 1 indicated that Cu(II), Ni(II)
and As(III) lower the conductivity of the electrolyte while temperature and H2 SO4 increase it, which is
in line with the literature [5,6].

3.2. Conductivity Model 2—Logarithmic Data with Terms of [H2 SO4 ]2 and T2
In order to avoid the skewness in Model 1 with untreated data, the second model was constructed
using logarithmic values of conductivity and this resulted in a more normally distributed histogram.
Terms with a high p-value were removed along with the two result series of the presumably
supersaturated electrolytes determined from Model 1. As before, the term squares T2 and [H2 SO4 ]2
were added to reduce the irregularities of the normal probability plot.
Equation (2) for the logarithm of the conductivity was compiled using unscaled coefficients. The
strongest combined effects based on low p-values were shown to be with H2 SO4 ¨T, H2 SO4 ¨As(III) and
T¨As(III); however, as with Model 1, the combined effects were minor compared to the single effects
of factors.

log10 pκq “ 2.17388 ´ 0.0023479 rCupIIqs ´ 0.0027733 rNipIIqs ´ 0.00073729 rAspIIIqs `

0.0037764 rH2 SO4 s ´ 1.0649 ˆ 10-5 rH2 SO4 srAspIIIqs ` 2.1627 ˆ 10-5 TrAspIIIqs ` (2)

8.8019 ˆ 10-6 rH2 SO4 sT ` 0.0051846 T ´ 6.9222 ˆ 10-6 rH2 SO4 s2 ´ 3.2506 ˆ 10-5 T2 ,
Minerals 2016, 6,
Model 2 59
was
regarded as reasonable and better with respect to Model 1, due to the improved 6 of 10
2 2 2
model validity and Q value, and Q was almost equal to R (Table 2). However, the condition number
4.456 was
4.456 was over
over3,3,asaswas
wasthe
thecondition
condition number
number of of Model
Model 1, but
1, but not not
overover 6, which
6, which would
would indicate
indicate a poor a
poor model. The condition number of this model was, however, lower
model. The condition number of this model was, however, lower than in Model 1.than in Model 1.

3.3. Conductivity
3.3. Conductivity Model
Model 3—Untreated Data without
3—Untreated Data without Arsenic
Arsenic
The third
The third conductivity
conductivity model
modelwaswasconstructed
constructedwithout
withoutthe theexperiments
experimentscontaining
containing arsenic
arsenic and
and it
it was
was found
found out
out thatthe
that thedata
datawas
wasquite
quitenormally
normallydistributed
distributed(Figure
(Figure5).
5).Scatter
Scatter plots
plots without
without arsenic
arsenic
did not
did not vary
vary remarkably
remarkably from from the
the corresponding plots with
corresponding plots with arsenic.
arsenic. Analogously, according to
Analogously, according to these
these
plots, Cu(II) and Ni(II) were shown to lower the conductivity while temperature
plots, Cu(II) and Ni(II) were shown to lower the conductivity while temperature and H2 SO4 were and H 2SO 4 were
shown to
shown to increase it. This
increase it. This model,
model, however, seems to
however, seems to have
have slightly
slightly better
better linearity
linearity in
in the
the relationships
relationships
between the factors and the response than that observed in Models 1 and 2
between the factors and the response than that observed in Models 1 and 2 which include thewhich include the effect
effect
of As(III). Thus, As(III) seems to cause non-linearity in the model and also affects the conductivity
of As(III). Thus, As(III) seems to cause non-linearity in the model and also affects the conductivity in in
more complex
more complex ways
ways than
than would
would be
be expected
expected from
from the
the literature
literature [5,6,10].
[5,6,10].

15
Count

10

0
453 488 523 558 593 628 663 698 733 768 803 838
mS/cm

5. Histogram
Figure 5. Histogram of ofconductivity
conductivityvalues
valuesofofsynthetic copper
synthetic electrorefining
copper electrolytes
electrorefining without
electrolytes the
without
effect of arsenic
the effect (for (for
of arsenic Model 3). 3).
Model

Model 33was
Model wasrefined
refined
likelike Model
Model 1 and1 compiled
and compiled according
according to unscaled
to unscaled coefficients.
coefficients. The
The combined
combined effect of H 2SO4·Ni(II) and H2SO4·T was shown to affect the conductivity value; however,
effect of H2 SO4 ¨ Ni(II) and H2 SO4 ¨ T was shown to affect the conductivity value; however, the single
the single parameters
parameters (Cu(II), H2(Cu(II), H2SOand
SO4 , Ni(II) 4, Ni(II) and T) had the biggest impact on conductivity. The sign
T) had the biggest impact on conductivity. The sign (˘) of
(±) of an individual variable or combined
an individual variable or combined effect effect of variables
of variables in in
thethe equationshould
equation shouldnot
notbebe interpreted
interpreted
individually, but as a combined effect of all variables that have an effect in the equation. Figure 6
shows the effect of Ni(II) and T on the conductivity according to Equation (3), which indicates
increased conductivity with increased temperature and decreased nickel concentration. Model
validity was good according to the summary of fit (Table 2).
 Figure 5. Histogram of conductivity values of synthetic copper electrorefining electrolytes without
the effect of arsenic (for Model 3).

Model 3 was refined like Model 1 and compiled according to unscaled coefficients. The
combined
Minerals 2016,effect
6, 59 of H2SO4·Ni(II) and H2SO4·T was shown to affect the conductivity value; however, 7 of 11
the single parameters (Cu(II), H2SO4, Ni(II) and T) had the biggest impact on conductivity. The sign
(±) of an individual variable or combined effect of variables in the equation should not be interpreted
individually, but
individually, butasasa acombined
combined effect of all
effect of variables thatthat
all variables havehave
an effect in thein
an effect equation. FigureFigure
the equation. 6 shows 6
the effect of Ni(II) and T on the conductivity according to Equation (3), which
shows the effect of Ni(II) and T on the conductivity according to Equation (3), which indicates indicates increased
conductivity
increased with increased
conductivity temperature
with increased and decreased and
temperature nickeldecreased
concentration.
nickelModel validity wasModel
concentration. good
according to the summary of fit (Table 2).
validity was good according to the summary of fit (Table 2).

κ “ 307.9 ´ 1.583 rCupIIqs ` 2.737 rNipIIqs ` 1.285 rH2 SO4 s ´ 0.02776 rH2 SO4 srNipIIqs ´
κ = 307.9 − 1.583 [Cu(II)] + 2.737 [Ni(II)] + 1.285 [H2SO4] − 0.02776 [H2SO4][Ni(II)] −
0.008774 (3)
rCupIIqsrH
0.008774 2 SO
[Cu(II)][H 4] 0.02087
4s ´
2SO [Ni(II)]T`+0.02919
rNipIIqsT
− 0.02087 [H22SO
0.02919rH SO44sT − 1.17
]T ´ 1.17 TT (3)

Figure 6. The
Figure 6. Theeffect
effectofoftemperature
temperature
andand Ni(II)
Ni(II) concentration
concentration on copper
on copper electrorefining
electrorefining electrolyte
electrolyte using
using Model 3, with [Cu(II)] = 50.51 g/L and [H 2SO4] = 182.842 g/L.
Model 3, with [Cu(II)] = 50.51 g/L and [H SO ] = 182.842 g/L.
2 4

3.4. Conductivity Model 4—Without Arsenic and with Terms of [H2 SO4 ]2 and T2
The fourth conductivity model was constructed without arsenic data and by adding extra terms
of [H2 SO4 ]2 and T2 , identical to Models 1 and 2.

κ “ 31.863 ´ 1.3594 rCupIIqs ` 1.835 rNipIIqs ` 2.9789 rH2 SO4 s ´ 0.022681 rH2 SO4 srNipIIqs ´
0.010403 rCupIIqsrH2 SO4 s ´ 0.021408 rNipIIqsT ` 0.02975 rH2 SO4 sT ` 2.7297 T ´ (4)
2 2
0.0044364 rH2 SO4 s ´ 0.032787 T

Both Models 3 and 4 were shown to be valid according to the summary of fit (Table 2), with model
validity being better than that of Model 3. Conversely, the condition number, 1.375, was better in
Model 3 when compared to the value of 4.322 in Model 4. In contrast, the condition number of the
unrefined design was 1.838.

3.5. Summary of the Models

The defined Equations (1)–(4) are relatively complex due to the interactions of the factors, and
thus the effects of the factors are impossible to directly see in the equations. The effects of Cu(II) and
H2 SO4 on electrolyte conductivity containing the median amount of As(III) and Ni(II) at medium
temperature, defined with Model 1, are presented in Figure 7a. Analogously, the effects of As and
temperature containing a high amount of Cu(II), a low amount of H2 SO4 and a medium amount of
Ni(II) are displayed in Figure 7b.
Figure 8 presents the measured and predicted electrolyte conductivity values. It can be seen that
the models predict the data with high correlation, with R2 varying from 0.9972 to 0.9985. In addition,
the R2 , Q2 , model validity and reproducibility values of the models are presented in Table 2.
 The defined Equations (1)–(4) are relatively complex due to the interactions of the factors, and
thus the effects of the factors are impossible to directly see in the equations. The effects of Cu(II) and
H2SO4 on electrolyte conductivity containing the median amount of As(III) and Ni(II) at medium
temperature, defined with Model 1, are presented in Figure 7a. Analogously, the effects of As and
temperature
Minerals 2016, 6, containing
59 a high amount of Cu(II), a low amount of H2SO4 and a medium amount 8 of of
11
Ni(II) are displayed in Figure 7b.

Minerals 2016, 6, 59 8 of 10
(a) (b)
temperature 7.seemed
Figure 7. The to be similar according to [12] and this (a)work. These comparisons are presented
Figure The effect
effect of Cu(II) and
of Cu(II) and HH2SO 4 concentrations as well as As(III) concentration
2 SO4 concentrations (a) as well as As(III) concentration and
and
in Figure 9, where
temperature (b)
temperature
the
(b) on equation
on copper by Devochkin
copper electrorefining et
electrorefining electrolyteal. has been
electrolyte (Model
(Model 1), corrected
1),with due
withTT==60 to
60˝°C, an error in the
[Ni(II)]== 10.102
C, [Ni(II)] 10.102g/L sign of
g/L
the Cu(II)
and concentration,
and [As(III)]
[As(III)] == 15
15 g/L (a)which
g/L (a) was noticed
and [Cu(II)]
and [Cu(II)] 60.6125when
== 60.6125 g/L, [H
g/L, [Htesting
2SO4] = the equation
162.519 g/L andwith their
[Ni(II)] parameters
= 10.102 g/L (b). and
2 SO4 ] = 162.519 g/L and [Ni(II)] = 10.102 g/L (b).
comparing the results to their measured values.
Figure 8 presents the measured and predicted electrolyte conductivity values. It can be seen that
the models predict the data with high correlation, with R2 varying from 0.9972 to 0.9985. In addition,
the R2, Q2, model validity and reproducibility values of the models are presented in Table 2.
The effects of temperature, Cu(II), As(III) and Ni(II) on conductivity are presented in
Figures 9–11. In addition to conductivity data calculated by Models 1–4, these figures present the
corresponding literature values and in Figure 10 five measured values are also shown. The equation
defined by Subbaiah and Das [8] was, however, not used in these comparisons, since it did not
reproduce the values they presented in their paper even when their own parameters were used. This
is probably due to the fact that their equation did not contain a Cu term, which possibly caused the
discrepancy as that error was at minimum at low Cu(II) concentrations.
Conductivity results obtained in this work were shown to be in good agreement with the
previous research work of Price and Davenport [6]. Nevertheless, arsenic was shown to affect the
conductivity slightly more and temperature less than determined by previous works [6,10].
Comparison of the results from this work to the equivalent results of Price and Davenport [6], Kern
and Chang [10] and Devochkin et al. [12] shows that there were some differences between them. The
Figure8.
Figure Observed(x)
8.Observed (x)versus
versuspredicted
predicted(y) (y)copper
copper electrolyte conductivity values using Models 1–4.
conductivity values were seen to be lower in [12] electrolyte
than in the conductivity values
other studies. using Models
Conversely, 1–4.
the effect of

The effects of temperature, Cu(II), As(III) and 160 g/l H2on

Ni(II) SO4conductivity are presented in Figures 9–11.
650
In addition to conductivity data calculated Modelby Models
2 (A) (B) 1–4, these figures present the corresponding
Price and Davenport (A) (B)
literature values and in Figure 600 10 five measured values
Devochkin et al., are also
corrected equation (A) shown.
(B) The equation defined by
Conductivity (mS/cm)

Subbaiah and Das [8] was, however, not used in these comparisons, since it did not reproduce the
550
values they presented in their paper even when their own parameters were used. This is probably due
to the fact that their equation did500 not contain a Cu term, which possibly caused the discrepancy as that
error was at minimum at low Cu(II) concentrations.
450
Conductivity results obtained in this work were shown to be in good agreement with the previous
research work of Price and Davenport 400 [6]. Nevertheless, arsenic was shown to affect the conductivity
slightly more and temperature less than determined by previous works [6,10]. Comparison of the
350
results from this work to the equivalent55results of 60 Price and 65 Davenport 70 [6], Kern and Chang [10] and
Devochkin et al. [12] shows that there were some differences T (°C) between them. The conductivity values
were seen to be lower in [12] than in the other studies. Conversely, the effect of temperature seemed to
Figure 9. Effects of Cu(II), Ni(II) and temperature according to Model 2 from this work compared to
be similar according to [12] and this work. These comparisons are presented in Figure 9, where the
results of Price and Davenport [6] and Devochkin et al. [12]. (A) 50 g/L Cu(II), 18 g/L Ni(II) and
equation by Devochkin et al. has been corrected due to an error in the sign of the Cu(II) concentration,
(B) 65 g/L Cu(II), 24 g/L Ni(II).
which was noticed when testing the equation with their parameters and comparing the results to their
measured values. 65 °C, 160 g/l H SO , 40 g/l Cu(II)
2 4

700 0 g/l Ni(II), this work, measured

0 g/l Ni(II), Model 2
680 10 g/l Ni(II), Model 2
20 g/l Ni(II), this work, measured
vity (mS/cm)

660
20 g/l Ni(II), Model 2
640 20 g/l Ni(II), Price and Davenport

620
600
Minerals 2016, 6, 59 9 of 11
Figure 8.
Figure 8. Observed
Observed (x)
(x) versus
versus predicted
predicted (y)
(y) copper
copper electrolyte
electrolyte conductivity
conductivity values
values using
using Models
Models 1–4.
1–4.

160 g/l
160 g/l H
H2SOSO4
2 4
650
650
Model 22 (A)
Model (A) (B)
(B)
Price and
Price and Davenport
Davenport (A) (A) (B)
(B)
600
600 Devochkin et al., corrected equation (A)
Devochkin et al., corrected equation (A) (B)
(B)

(mS/cm)
Conductivity(mS/cm)
550
550

500

Conductivity
500

450
450

400
400

350
350
55
55 60
60 65
65 70
70
T (°C)
T (°C)

Figure 9.
Figure 9. Effects
9. Effects
Effectsofof Cu(II),
ofCu(II), Ni(II)
Cu(II),Ni(II) and
Ni(II)and temperature
andtemperature according
temperatureaccording
according toto
to Model
Model
Model 22 from
from this
this
2 from work
work
this workcompared
compared
compared to
to
results
results
to of of
of
results Price
Price
Priceand
andandDavenport
Davenport
Davenport[6][6] and
[6]and Devochkin
andDevochkin
Devochkinetetetal.
al. [12].
al.[12]. (A)
[12]. (A) 50
(A) 50 g/L Cu(II),
50 g/L
g/L Cu(II), 18
Cu(II), 18 g/L Ni(II)
g/L
18 g/L Ni(II) and
Ni(II) and
(B) 65
(B) 65 g/L
g/L Cu(II),
Cu(II), 24
24 g/L
g/L Ni(II).
Ni(II).
g/L Cu(II), 24 g/L Ni(II).

65 °C,
65 °C, 160
160 g/l
g/l H
H2SO
SO4,, 40
40 g/l
g/l Cu(II)
Cu(II)
2 4
700
700 00 g/l
g/l Ni(II),
Ni(II), this
this work,
work, measured
measured

680 00 g/l
g/l Ni(II),
Ni(II), Model
Model 22
680 10 g/l
10 g/l Ni(II),
Ni(II), Model
Model 22
20 g/l
g/l Ni(II),
Ni(II), this
this work,
work, measured
measured
(mS/cm)

660 20
Conductivity(mS/cm)

660 20 g/l
g/l Ni(II),
Ni(II), Model
Model 22
20
640
640 20 g/l Ni(II), Price and Davenport
20 g/l Ni(II), Price and Davenport
620
620
Conductivity

600
600
580
580
560
560
540
540
520
520
500
500
00 55 10
10 15
15 20
20 25
25 30
30 35
35 40
40
As(III) (g/l)
As(III) (g/l)

Figure 10.
Figure 10. Effects
Effects of
Effects of Ni(II)
of Ni(II) and
Ni(II) and As(III)
and As(III) according
As(III) according to
according to measured
to measured values
measured values and
values and Model
Model 222 from
and Model from this work
from this work
compared
Minerals to results
2016, 6, 59
compared to results of
of Price
Price and
and Davenport
Davenport [6].
[6]. 9 of 10

65 °C, 160 g/l H2SO4, 40 g/l Cu(II), 10 g/l Ni(II)

600
Model 1
Model 2
590 Model 3
Conductivity (mS/cm)

Model 4
Price and Davenport
580

570

560

550

0 5 10 15 20 25 30
As(III) (g/l)

11.Effects
Figure 11. Effectsof of As(III)
As(III) on copper
on copper electrorefining
electrorefining electrolyte
electrolyte conductivity
conductivity defineddefined using
using Models
Models
1–4 from1–4
thisfrom
workthis
andwork and compared
compared to valuestoofvalues of Price
Price and and Davenport
Davenport [6]. [6].

Table 3 presents a comparison of values calculated with Models 1–4 and the model of Price and
Davenport [6], Devochkin et al. [12] and measurements of Kern and Chang [10]. The model of
Devochkin et al. was designed to determine conductivity from electrolytes without As(III) and with
a constant Cu(II)/Ni(II) ratio; therefore, for that reason, their model was not used with all parameter
combinations. The conductivity value determined using their model was lower than the other
equivalent values, as can be observed in Figure 7. The values predicted by Models 1–4 were in good
Minerals 2016, 6, 59 10 of 11

Table 3 presents a comparison of values calculated with Models 1–4 and the model of Price
and Davenport [6], Devochkin et al. [12] and measurements of Kern and Chang [10]. The model
of Devochkin et al. was designed to determine conductivity from electrolytes without As(III) and
with a constant Cu(II)/Ni(II) ratio; therefore, for that reason, their model was not used with all
parameter combinations. The conductivity value determined using their model was lower than the
other equivalent values, as can be observed in Figure 7. The values predicted by Models 1–4 were in
good agreement with the values measured (Figure 8).

Table 3. Comparison of values defined with Models 1–4 to equivalent values from Price and
Davenport [6], Devochkin et al. [12] and to measured values from Kern and Chang [10].

κ (mS/cm)
Concentration (g/L)
T (˝ C) Model Price and Kern and Devochkin
Davenport Chang et al.
H2 SO4 Cu(II) Ni(II) As(III) 1 2 3 4
55 135 35 0 0 538.4 534.0 536.9 528.3 527.3 530.7 -
55 135 35 30 0 453.4 440.9 472.1 456.1 458.4 444.0 -
55 135 35 0 30 504.8 498.9 - - 516.7 523.1 -
55 135 35 0 40 493.6 487.7 - - 513.2 519.4 -
55 150 50 18 0 478.0 475.0 485.8 479.9 477.8 - 446.13

Furthermore, it can be seen that the values defined using the models detailed in this work also
show a good correlation with both Kern and Chang’s results [10] and Price and Davenport (who
reported a good agreement the results of Kern and Chang) [6]. According to this research, arsenic
decreases conductivity and an improved model can be constructed.

4. Conclusions
In this work the conductivity of the copper electrorefining electrolyte was investigated as a
function of temperature (50–70 ˝ C) and concentrations of copper (Cu(II), 40–60 g/L), nickel (Ni(II),
0–20 g/L), arsenic (As(III), 0–30 g/L) and sulfuric acid (160–220 g/L). In total, 165 different
combinations of these factors were studied.
The measured data showed that conductivity was increased by a decrease in Cu(II), Ni(II) and
As(III) concentration and increased with increasing temperature and acidity. As(III) appeared to
affect the conductivity in more complicated ways than would be expected based on findings from
the literature. In addition, it was observed that the untreated measured data was shown to be
slightly skewed.
The results were treated using factorial analysis, and as a result four different electrolyte
conductivity models were created. Conductivity was measured reliably, and thus the models from the
conductivity results were constructed directly. Combined effects were also detected, but the effects
were minor compared to the effects of single factors.
Model 1 was constructed from untreated conductivity data with terms of [H2 SO4 ]2 and T2 . Model
2 was constructed from logarithmic values of conductivity, also with additional terms of [H2 SO4 ]2 and
T2 . Models 3 and 4 were constructed from untreated data by neglecting the measurement series with
arsenic, and Model 4 with additional terms of [H2 SO4 ]2 and T2 . The four models constructed were all
valid and had high correlation coefficients. In addition, the reproducibility was good, and the models
did not suffer from a lack of fit. The most accurate models based on the best R2 , Q2 and model validity
were Model 2 and Model 4 (Figure 8 and Table 2).
Overall, this work provides improved models for copper electrolyte conductivity both in the
presence and absence of arsenic. In particular, As(III) is shown to cause slight non-linearity and
decrease conductivity more than previously reported, whereas temperature is shown to affect the
electrolyte conductivity slightly less.
Minerals 2016, 6, 59 11 of 11

Acknowledgments: This research has been performed within the SIMP (System Integrated Metal Production)
project of FIMECC (Finnish Metals and Engineering Competence Cluster Ltd. (Tampere, Finland)).
Author Contributions: Taina Kalliomäki performed the experiments and analyzed the data as part of her
Master’s thesis; Jari Aromaa counseled in experimental and theoretical matters as an instructor of the thesis;
Mari Lundström and Taina Kalliomäki wrote the paper. All authors have read and approved the manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Davenport, W.G.; King, M.; Schlesinger, M.; Biswas, A.K. Overview. In Extractive Metallurgy of Copper, 4th ed.;
Elsevier Science Ltd.: Amsterdam, The Netherlands, 2002.
2. Wraith, A.E.; Mackey, P.J.; Jones, R.P. Origins of electrorefining: Birth of the technology and the world’s
first commercial electrofinery. In Proceedings of Copper 2013; The Chilean Institute of Mining Engineers:
Santiago, Chile, 2013.
3. Moats, M.S.; Hiskey, J.B. How anodes passivate in copper electrorefining. In The Copper 2010-Proceedings;
GDMB: Hamburg, Germany, 2010; Volume 4.
4. Davenport, W.G.; King, M.; Schlesinger, M.; Biswas, A.K.; Robinson, T. Electrolytic Refining. In Extractive
Metallurgy of Copper, 4th ed.; Elsevier Science Ltd.: Amsterdam, The Netherlands, 2002; pp. 265–288.
5. Price, D.C.; Davenport, W.G. Densities, electrical conductivities and viscosities of CuSO4 /H2 SO4 solutions
in the range of modern electrorefining and electrowinning electrolytes. Metall. Trans. B 1980, 11, 159–163.
[CrossRef]
6. Price, D.C.; Davenport, W.G. Physico-chemical properties of copper electrorefining and electrowinning
electrolytes. Metall. Trans. B 1981, 12, 639–643. [CrossRef]
7. Moats, M.S.; Hiskey, J.B.; Collins, D.W. The effect of copper, acid, and temperature on the diffusion coefficient
of cupric ions in simulated electrorefining electrolytes. Hydrometallurgy 2000, 56, 255–268. [CrossRef]
8. Subbaiah, T.; Das, S.C. Physico-chemical properties of copper electrolytes. Metall. Trans. B 1989, 20, 375–380.
[CrossRef]
9. Jarjoura, G.; Muinonen, M.; Kipouros, G.J. Physicochemical properties of nickel copper sulfate solutions.
Can. Metall. Q. 2003, 42, 281–288. [CrossRef]
10. Kern, E.F.; Chang, M.Y. Conductivity of copper refining electrolytes. In Proceedings of the 41st General
Meeting of the American Electrochemical Society, Baltimore, MD, USA, 28 April 1922.
11. Skowronski, S.; Reinoso, E.A. The specific resistivity of copper refining electrolytes and method of calculation.
In Proceedings of the 51st General Meeting of the American Electrochemical Society, Philadelphia, PA, USA,
30 April 1927.
12. Devochkin, A.I.; Kuzmina, I.S.; Salimzhanova, E.V.; Petukhova, L.I. The study of sulfate copper electrolyte
physicochemical properties depending on its components and temperature. Tsvetnye Met. 2015, 2015, 67–71.
[CrossRef]
13. Zhang, X.; Hu, Y.; Peng, X.; Yue, W. Conductivities of several ternary electrolyte solutions and their binary
subsystems at 293.15, 298.15, and 303.15 K. J. Solut. Chem. 2009, 38, 1295–1306. [CrossRef]
14. Eriksson, L.; Johansson, E.; Kettaneh-Wold, N.; Wikström, C.; Wold, S. Design of Experiments: Principles and
Applications; MKS Umetrics AB: Malmö, Sweden, 2008.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).