metals

Article
Effect of Aqueous Media on the Recovery of
Scandium by Selective Precipitation
Bengi Yagmurlu 1,2, * ID
, Carsten Dittrich 1 and Bernd Friedrich 2 ID

1 MEAB Chemie Technik GmBH, 52068 Aachen, Germany; [email protected]

2 RWTH Aachen, IME Institute of Process Metallurgy and Metal Recycling, 52056 Aachen, Germany;
[email protected]
* Correspondence: [email protected]; Tel.: +49-1575-4954583




Received: 9 April 2018; Accepted: 1 May 2018; Published: 3 May 2018


Abstract: This research presents a novel precipitation method for scandium (Sc) concentrate refining
from bauxite residue leachates and the effect of aqueous media on this triple-stage successive
precipitation process. The precipitation pattern and the precipitation behavior of the constituent
elements was investigated using different precipitation agents in three major mineral acid media,
namely, H2 SO4 , HNO3 , and HCl in a comparative manner. Experimental investigations showed
behavioral similarities between HNO3 and HCl media, while H2 SO4 media was different from them
because of the nature of the formed complexes. NH4 OH was found to be the best precipitation agent
in every leaching media to remove Fe(III) with low Sc co-precipitation. To limit Sc loss from the
system, Fe(III) removal was divided into two steps, leading to more than 90% of Fe(III) removal at
the end of the process. Phosphate concentrates were produced in the final step of the precipitation
process with dibasic phosphates which have a strong affinity towards Sc. Concentrates containing
more than 50% of ScPO4 were produced in each case from the solutions after Fe(III) removal, as
described. A flow diagram of the selective precipitation process is proposed for these three mineral
acid media with their characteristic parameters.

Keywords: bauxite residue; red mud; hydrometallurgy; recovery; scandium; precipitation

1. Introduction
The recent agreements and climate accords in reducing carbon emission and specified deadlines
for automotive industries have placed light metals and alloys under the spotlight [1]. One of the
reasons is the direct relation between a vehicle weight and its energy consumption. Scandium (Sc) is
used as a tuning metal especially for aluminum alloys, which makes it one of the promising candidates
for light-weight alloys [2]. It is, however, an extremely expensive element for widespread application
in industrial usage at the moment [3]. As aluminum alloys with improved strength, thermal resistance,
and weldability can be achieved with minor additions of Sc, improved oxygen-ion conductivity can
also be attained in solid oxide fuel cells [4–6]. Hence, this metal was classified recently as a critical
metal for the future, leading to a steep increase in its demand despite its price [7].
Unfortunately, Sc is widely dispersed in nature and generally has to be extracted from secondary
raw materials or as a by-product of uranium, nickel laterite, or titanium pigment processing. Bauxite
residue (i.e., red mud) is the by-product obtained through the Bayer Process, yielding approximately
four billion tons, with a previously reported annual production of 160 million tons [8,9]. This alkaline
waste can be considered as a valuable resource because of its metal content (Fe, Al, Ti, Sc, Rare earth
elements (REEs), etc.). Therefore, the complete or partial valorization of bauxite residues (BR) has
lately been of great interest [10–13].

Metals 2018, 8, 314; doi:10.3390/met8050314 www.mdpi.com/journal/metals

Metals 2018, 8, 314 2 of 13

Previously, complete or partial recovery of Sc from bauxite residues was reported to be achieved
mainly by solvent extraction, ion exchange, or the combination of these two techniques, as a result of
its low concentration in the leachates [14–20]. Zhang et al. recovered 91% of Sc from bauxite residue
leachates by inorganic metal(IV)–phosphate ion exchangers, although Fe(III) was found to be an
interfering ion in this process [20]. In another study, a newly developed supported ionic liquid phase
(SILP) achieved almost complete Sc extraction, while showing a decreased efficiency in the presence of
Fe(III) [14]. In all of these hydrometallurgical operations, the co-extraction of Fe, Al, and Ti became a
problem, and intensive purification was required to produce a high-quality product.
In our previous work, a three-staged precipitation process was designed using a sulfuric acid
media based on a selective Fe removal step by NH4 OH, since Fe is the most problematic element
during Sc processing [21,22]. This was successively followed by selective Sc phosphate precipitation by
(NH4 )2 HPO4 . As a result of this precipitation route, a Sc phosphate concentrate containing 65% Sc was
synthesized from impure synthetic bauxite residue leach solutions. Nevertheless, the processing route
must be tailored in relation to the geological presence of the bauxite residue, depending on different
minerology and association of the phases. Hence, different mineral acids other than H2 SO4 , such as
HNO3 and HCl, were also tried for leaching bauxite residues.
In order to cover a wider range of bauxite residues as well as other waste-generating processes,
such as Ti-pigment and Ni laterite production, the design of selective Sc precipitation route for bauxite
residues has to be adapted to different mineral acid media. Thus, this paper investigates the effect of
different aqueous media on the recovery of Sc by a selective precipitation method.

2. Materials and Methods

A bauxite residue sample was obtained from Aluminium of Greece, subjected to lithium borate
fusion, and analyzed using inductively coupled plasma mass spectrometry, ICP-MS/AAS, as shown
in Table 1.

Table 1. Chemical composition of the bauxite residue (LOI: Loss of ignition).

Major Compounds wt. % Minor Compounds ppm

Fe 29.6 La 110
Al 8.6 Ce 380
Ca 8.3 Sc 120
Si 3.3 Nd 100
Ti 2.6 Y 80
Na 2.8
Others 32.1
LOI 12.7
LOI: Loss on ignition

Synthetic leach solutions were prepared considering pre-treatments before performing the
selective precipitation. The expected leachate is the pregnant leach solutions (PLS) after the major part
of Fe, Al, and Ti are recovered from the solution by both pyrometallurgical and hydrometallurgical
methods. Hence, the synthetic solution mentioned in this study is predictive of the formation of the real
PLS. The Sc and REEs concentrations were arranged to be higher than in the real PLS to clearly observe
the precipitation behaviors of these elements. Additionally, the selective precipitation process was also
tested directly on the real solutions, obtaining similar results as those observed in this study [21,23].
In addition to the major impurities present in the bauxite residue (e.g., Fe and Al), Nd and Y are
also considered representative elements of light and heavy REEs within the synthetic PLS, since they
have similar chemical properties as same sub-groups of REEs.
Synthetic solutions in chloride media were prepared by adding the required amount of
reagent-grade FeCl3 ·6H2 O, AlCl3 , Sc2 (SO4 )3 ·5H2 O, NdCl3 ·6H2 O, and YCl3 ·6H2 O. In a similar manner,
reagent-grade Fe(NO3 )3 ·9H2 O, Al2 (SO4 )3 ·18H2 O, Sc2 (SO4 )3 ·5H2 O, NdCl3 ·6H2 O, and YCl3 ·6H2 O were
Metals 2018, 8, 314 3 of 13

used to synthesize the solution in nitrate media. Sulfate and chloride salts were first precipitated as
hydroxides and washed before being converted into the necessary forms, to avoid unwanted sulfate or
chloride ions which can affect the precipitation yields by promoting complex formations in the aqueous
solution. All precipitation solutions were prepared from reagent-grade salts. The concentrations of the
precipitation agents were 12.5 wt. % for CaCO3 (limestone) slurry, 1 mol/L for NaOH, NH4 OH, and
KOH, and 1 mol/L for K2 HPO4 , (NH4 )2 HPO4 , and Na2 HPO4 .
In order to have comparative precipitation results, a composition similar to the one used in the
sulfate media was chosen. The pH of both systems was set between 1.2 and 1.4 with the addition of the
necessary amounts of HCl or HNO3 . The composition of the synthetic solutions is presented in Table 2.

Table 2. Composition of the synthetic solutions in HNO3 and HCl media.

Composition in HNO3 Media (mg/L) Composition in HCl Media (mg/L)

Al 340 348
Fe(III) 312 290
Sc 86 78
Y 108 96
Nd 101 105

The precipitation agents mentioned were carefully added using a precision burette into 50 mL
of the synthetic PLS while monitoring pH and temperature. All experiments presented in this study
were done at room temperature. For hydroxide precipitation, the agents were added until the target
pH was attained under mild agitation, to reach homogeneity in the solution and to prevent local
pH differences.
Precipitation solutions containing dibasic phosphates were prepared as 1 mol/L and added into
the leach solutions starting with a stoichiometric amount, considering only scandium precipitation.
In each step, the amount added was doubled until reaching 20 times of the stoichiometric amount.
The resulted suspension for each case was then stabilized and homogenized at a given pH
and temperature for 2 h and subsequently filtered through fine filter paper via suction filtration.
The separated solid residue was washed with distilled water and dried at 110 ◦ C for 24 h. Both filtered
solutions and solid residues were assayed.
The concentrations of the constituent ions of iron (Fe), aluminum (Al), scandium (Sc), yttrium (Y),
and neodymium (Nd) were determined by microwave plasma optical emission spectroscopy (Agilent
MP-AES 4100, Mulgrave, VIC, Australia). Each sample was prepared by adding 100 µL of cesium
ionization buffer and 500 µL of ultrapure concentrated HNO3 to 10 mL of solution. Quantitative
analyses were performed at 371.993 nm, 396.152 nm, 361.383 nm, 371.029 nm, and 430.358 nm,
corresponding to the spectral emission lines for Fe, Al, Sc, Y, and Nd respectively.
The pH measurements were performed using a WTW ProfiLine pH 197 series pH-meter with a
Sentix 81 precision electrode. The pH meter was calibrated with standard technical buffering solutions
at pH 2.00, 4.01, and 7.00 to achieve maximum sensitivity in pH measurements.

3. Results
We previously reported that more than 90% of Fe can be removed with negligible amount of
Sc loss from sulfuric acid-based solutions by simply adding NH4 OH in a dual-staged precipitation
process [22]. Furthermore, with a successive precipitation route which combines both hydroxide
precipitation and phosphate precipitation, a Sc phosphate concentrate, which is easier to process, can
be synthesized from sulfate-based aqueous solutions. Since Sc-containing liquors can also exist in
chloride or nitrate media, the effect of those aqueous media on precipitation should be investigated.
By this way, the successive precipitation process can be adapted to other major mineral acid media.
Metals 2018, 8, 314 4 of 13

3.1. Precipitation
Metals 2018, 8, x in HNO
FOR Media
PEER3REVIEW 4 of 13

3.1.1.3.1.
Fe Precipitation
Removal Step in HNO3 Media
In all recovery and purification operations regarding scandium, iron was reported to be one of the
3.1.1. Fe Removal Step
most problematic elements. Thus, to propose an easier route to process scandium, iron content should
be minimizedIn all in
recovery and purification
the solution beforehand. operations regardingthat
We determined scandium, iron route
the easiest was reported to be onewas
to this purpose of the
the most problematic elements. Thus, to propose an easier route to process scandium, iron content
hydroxide precipitation. Limestone slurry, sodium hydroxide, potassium hydroxide, and ammonium
should be minimized in the solution beforehand. We determined that the easiest route to this purpose
hydroxide were chosen as precipitation solutions, and a wide range of pH values was examined with
was the hydroxide precipitation. Limestone slurry, sodium hydroxide, potassium hydroxide, and
the addition
ammonium of these hydroxide
hydroxide donors.
were chosen as precipitation solutions, and a wide range of pH values was
The precipitation trends
examined with the addition of theseof the hydroxide
constituent elements as hydroxides are shown in Figure 1.
donors.
The precipitation
The precipitation trends of the constituent elementsslurry
of Fe(III) with the addition of limestone took place
as hydroxides areinshown
the pHin range
Figurebetween
1. The 2.0
and 3.5. As it can of
precipitation beFe(III)
seen from Figure
with the 1a, when
addition Fe(III) precipitation
of limestone wasintriggered,
slurry took place the pH range thebetween
co-precipitation
2.0
of theand 3.5. As
other it can be seen
components from
was Figure25–30%.
around 1a, when Fe(III)
In the precipitation
case of NaOH wasand
triggered,
KOH the co-precipitation
addition to the system
of the
(Figure 1b,cother components
respectively), was around
a more 25–30%.
distinctive In the
Fe(III) case of NaOH
precipitation wasand KOH addition
observed. toco-precipitation
Yet, the the system
(Figure 1b,c respectively), a more distinctive Fe(III) precipitation
levels were relatively high, between 10% and 20%, when more than 90% of Fe(III) was was observed. Yet, theremoved.
co-
precipitation levels were relatively high, between 10% and 20%, when more than 90% of Fe(III) was
The addition of NH4 OH produced similar results as those described in the sulfate system. A distinctive
removed. The addition of NH4OH produced similar results as those described in the sulfate system.
Fe(III) precipitation resulting in low co-precipitation levels of the other components was observed
A distinctive Fe(III) precipitation resulting in low co-precipitation levels of the other components was
(Figure 1d). When
observed (Figure more
1d). than
When95%moreofthan
Fe was
95% removed from thefrom
of Fe was removed solution, less than
the solution, less5%
thanof 5%
Sc and
of ScREEs
precipitated. In all cases, the precipitation order was Fe(III) > Sc > Al > Y ≈
and REEs precipitated. In all cases, the precipitation order was Fe(III) > Sc > Al > Y ≈ Nd.Nd.

100 100
Y Y
a) Fe
b) Fe
Sc Sc
80 Nd 80 Nd
Al Al
Precipitation Efficiency (%)

Precipitation Efficiency (%)

60 60

40
40

20
20

0
0
1 2 3 4 5 6
1 2 3 4 5 6
pH
pH
100 100
Y Y
c) Fe d) Fe
Sc Sc
80 Nd 80 Nd
Al Al
Precipitation Efficiency (%)

Precipitation Efficiency (%)

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 8 1 2 3 4 5 6
pH pH

Figure 1. Precipitation behavior of Fe(III), Al, Sc, Nd, and Y in HNO3 media at different pH values
Figure 1. Precipitation behavior of Fe(III), Al, Sc, Nd, and Y in HNO3 media at different pH values
with the addition of (a) limestone; (b) NaOH; (c) KOH; (d) NH4OH.
with the addition of (a) limestone; (b) NaOH; (c) KOH; (d) NH4 OH.
In this removal step, it is desirable to remove Fe from the system with minimum co-precipitation
In thisrecoverable
of the removal step, it is desirable
elements; therefore,toselectivity
remove Fe from
was the system
an essential with minimum
parameter. co-precipitation
The selectivity of the
of theprecipitation
recoverable is elements;
calculated by Equationselectivity
therefore, (1) given below:
was an essential parameter. The selectivity of the
precipitation is calculated by Equation (1) given
C below: DA
DA , B = and S A/ B =
prec
(1)
CC
prec
aq DD
B
D A,B = and S A/B = A (1)
Caq DB
Metals 2018, 8, 314 5 of 13

where DA or DB is the distribution coefficient of the mentioned element, Cprec is the concentration of
the element in the precipitate, Caq is the concentration of the element in the aqueous solution after
precipitation, SA/B is the selectivity coefficient which indicates the selectivity of A over B.
The precipitation yields of the elements and the selectivity of Fe over Sc can be found in Table 3,
when Fe precipitation reach 70% and 90%, respectively. Table 3a shows that the best candidates
to remove Fe in HNO3 media are NaOH and NH4 OH, since the corresponding SFe/Sc values are
far superior to those of the other hydroxide donors. Nevertheless, the co-precipitation levels of all
elements abruptly increased in the case of NaOH upon further addition to remove 90% of Fe(III). Thus,
NH4 OH showed an astonishing performance considering both the co-precipitation levels of the other
elements and the SFe/Sc . Once that more than 95% of Fe present in the system was removed, around 4%
of Sc and 1% of the other elements were precipitated with a remarkable SFe/Sc of 585. The most logical
explanation of this low co-precipitation levels is the occurrence of an hexamine scandium complex
upon addition of NH4 OH [24,25]. Hence, Fe precipitation was triggered, whereas Sc remained in a
complex form, which prevented the co-precipitation of Sc between these pH ranges.
The main difference of the nitrate-based aqueous solution compared to the sulfate-based one regards
the pH ranges of the precipitation. Although it was found that the precipitation of Fe was triggered at
pH values between 2.5 to 4.0 in sulfate media, a similar precipitation level was observed at a lower pH,
between 2.0 to 3.0. It is known that sulfate ion form inner-sphere complexes, while nitrate complexes
can be classified as forming outer-sphere complexes [26,27]. Consequently, more OH− ions have to be
released to disrupt inner-sphere complexes, since they have lower Gibbs Energy (∆G) in that state.

Table 3. (a) Critical pH values for 70% Fe(III) removal from the system by hydroxide precipitation
and precipitation % of the constituent ions with selectivity of Fe over Sc; (b) Critical pH values to
obtain above 90% Fe(III) removal from the system by hydroxide precipitation and precipitation % of
the constituent ions with selectivity of Fe over Sc.

(a)
Precipitation Agent pH Fe (%) Sc (%) Al (%) Y (%) Nd (%) SFe/Sc
Limestone 2.95 76.6 26.8 25.1 28.1 24.2 9
NaOH 2.78 72.3 1.1 3.3 8.1 6.8 235
KOH 2.74 69.7 10.4 13.1 21.9 15.2 20
NH4 OH (aq) 2.87 71.5 2.4 0.5 0.5 0.5 102
(b)
Precipitation Agent pH Fe (%) Sc (%) Al (%) Y (%) Nd (%) SFe/Sc
Limestone 3.05 93.6 27.6 25.9 29.4 24.2 38
NaOH 3.00 88.9 11.0 5.1 9.3 7.4 65
KOH 2.91 93.4 13.3 15.3 23.5 17.4 92
NH4 OH (aq) 3.10 96.5 4.5 0.9 0.6 1.3 585

3.1.2. Phosphate Precipitation

In previous studies, it was shown that there is a strong affinity of Sc and REEs for PO4 3−
ion [22,28,29]. In leach solutions with sulfate media, addition of dibasic phosphate resulted in a
selective precipitation of both Fe(III) and Sc from the system. In light of this, dibasic phosphate
solutions were tested with the intention of recovering Sc from the PLS. Addition of three different
dibasic phosphate solutions and the resulting precipitation patterns of the constituent elements are
shown in Figure 2. Precipitation solutions were added starting from the stoichiometric amount
considering only Sc precipitation, and the amount introduced was increased in each step until reaching
an amount 20 times greater than the stoichiometric value.
In all cases, similar precipitation trends were observed. Upon addition of the precipitant solution,
immediate precipitation of Fe and Sc was initiated. More than 90% of Sc and Fe precipitation occurred
at a pH value around 2.5. Since dibasic phosphates release OH− into the system, an uncontrolled rise of
pH would lead to the precipitation of the constituent elements as hydroxides, which will deteriorate the
Metals 2018, 8, 314 6 of 13

selectivity towards
Metals 2018, 8, x FORSc recovery.
PEER REVIEW It is important to note that, while high levels of co-precipitation 6 of 13of the

other elements were detected with Na2 HPO4 addition, corresponding approximately to 40–50%, it was
levels of co-precipitation of the other elements were detected with Na2HPO4 addition, corresponding
found that they were limited between 20% and 25% when K2 HPO4 or (NH4 )2 HPO4 were introduced
approximately to 40–50%, it was found that they were limited between 20% and 25% when K2HPO4
into or
the(NH
solution as phosphate donors. In all cases, the additions yielded to the precipitation order
4)2HPO4 were introduced into the solution as phosphate donors. In all cases, the additions
Fe(III)
yielded to the >precipitation
= Sc > Al Y = Nd. Theorderprecipitation
Fe(III) = Sc > efficiencies
Al > Y = Nd. upon the addition
The precipitation of dibasic
efficiencies phosphate
upon the
precipitation
addition of solutions can be found
dibasic phosphate in Tablesolutions
precipitation 4. can be found in Table 4.

100 100
Y Y
a) Fe b) Fe
Sc Sc
80 Nd 80 Nd
Al Al
Precipitation Efficiency (%)

Precipitation Efficiency (%)

60 60

40 40

20 20

0 0
1 2 3 1 2 3

pH pH
100
Y
c) Fe
Sc
80 Nd
Al
Precipitation Efficiency (%)

60

40

20

0
1 2 3

pH

Figure 2. Precipitation of the elements in HNO3 media with the addition of (a) Na2HPO4; (b) K2HPO4
Figure 2. Precipitation of the elements in HNO3 media with the addition of (a) Na2 HPO4 ; (b) K2 HPO4
and (c) (NH4)2HPO4.
and (c) (NH4 )2 HPO4.
Table 4. Critical pH values for above 90% Sc recovery from the solution system by dibasic phosphate
4. Critical pH
Tableprecipitation and values for above
precipitation 90% Scofrecovery
efficiencies from ions.
the constituent the solution system by dibasic phosphate
precipitation and precipitation efficiencies of the constituent ions.
Precipitation Agent pH Fe(III) (%) Sc (%) Al (%) Y (%) Nd (%)
Na2HPO4 2.36 90.8 90.2 44.8 39.6 54.3
Precipitation Agent pH Fe(III) (%) Sc (%) Al (%) Y (%) Nd (%)
K2HPO4 2.24 91.2 91.3 25.1 24.8 24.9
Na2 HPO 4 4)2HPO4
(NH 2.36 2.30 90.8
89.3 90.2
90.2 44.8
36.3 22.0 39.631.2 54.3
K2 HPO4 2.24 91.2 91.3 25.1 24.8 24.9
(NH4 )2 HPO4 2.30 89.3 90.2 36.3 22.0 31.2
3.2. Precipitation in HCl Media

3.2. Precipitation in HCl

3.2.1. Fe Removal Media
Step
The same precipitation
3.2.1. Fe Removal Step solutions for hydroxide precipitation were further tested in HCl media,
and the results are summarized in Figure 3. Similar precipitation routes and behaviors were observed
The
as in same precipitation
the case of HNO3 media. solutions foraddition
While the hydroxide precipitation
of limestone slurrywere further
resulted in thetested in HCl media,
co-precipitation
of all
and the other are
results elements with Fe(III),
summarized more3.distinctive
in Figure cases were achieved
Similar precipitation routes andwhen NaOH, KOH,
behaviors and
were observed
NH OH were used. As in the previous case, with the addition of NH OH, the lowest co-precipitation
as in the case of HNO3 media. While the addition of limestone slurry resulted in the co-precipitation of all
4 4

otherwas achieved,
elements with yetFe(III),
Sc precipitation was observed
more distinctive to beachieved
cases were the greatest among
when the other
NaOH, KOH,elements
and NHwhen
4 OH were
the Fe content in the solution was minimized. In all cases, the precipitation order was Fe(III) > Sc > Al
used. As in the previous case, with the addition of NH4 OH, the lowest co-precipitation was achieved, yet
> Y = Nd.
Sc precipitation was observed to be the greatest among the other elements when the Fe content in the
The precipitation percentages of the elements in the solution can be found in Table 5, when
solution was minimized. In all cases, the precipitation order was Fe(III) > Sc > Al > Y = Nd.
Fe(III) was removed at 70% and 90%, respectively. The major difference in the precipitation behavior
The precipitation percentages of the elements in the solution can be found in Table 5, when
Fe(III) was removed at 70% and 90%, respectively. The major difference in the precipitation behavior
Metals 2018, 8, 314 7 of 13

Metals 2018, 8, x FOR PEER REVIEW 7 of 13

between HNO3 media and HCl media regarded the co-precipitation levels of Sc. It was previously
discussed
between thatHNO both in nitrate and in sulfate media, Sc remained in the solution, while Fe(III) was
3 media and HCl media regarded the co-precipitation levels of Sc. It was previously
almost completely
discussed takeninout
that both fromand
nitrate theinsolution. With the
sulfate media, addition in
Sc remained of the
NHsolution,
4 OH, 70% of Fe(III)
while Fe(III) as well as
was
9% ofalmost
Sc were precipitated
completely takenat pH
out values
from around 2.8,
the solution. Withwhile Sc precipitation
the addition of NH4OH, remained at 13%
70% of Fe(III) as once
well 93%
of Fe(III)
as 9%wasof Scremoved from the at
were precipitated solution
pH valuesat aaround
pH value around
2.8, while Sc 3.0. The interaction
precipitation remained between Fe3+ and
at 13% once
−
Cl ions was the main reason why the precipitation level was limited, and the co-precipitationFelevel
93% of Fe(III) was removed from the solution at a pH value around 3.0. The interaction between 3+
of
and Cl − ions was the main reason why the precipitation level was limited, and the co-precipitation
Sc increased when Fe started to precipitate. This interaction shifted the precipitation range of Fe(III)
levelprompting
slightly, of Sc increased when Fe started
Sc precipitation in to precipitate.
the same range.This interaction shifted the precipitation range of
Fe(III) slightly, prompting Sc precipitation in the same range.
Still, the precipitation procedure to synthesize a scandium concentrate is applicable even in this
Still, the precipitation procedure to synthesize a scandium concentrate is applicable even in this
case. Although the selectivity of Fe over Sc was found to be 24 when the precipitation of Fe reached
case. Although the selectivity of Fe over Sc was found to be 24 when the precipitation of Fe reached
70%,70%,
as theas precipitation
the precipitationprogressed,
progressed,a avalue
valueof
of 100 wasreached
100 was reachedwith
withthethe addition
addition of NH
of NH 4OH.4 OH.

100 100
Y Y
Fe Fe
a) Sc b) Sc
80 Nd 80 Nd
Al Al
Precipitation Efficiency (%)

Precipitation Efficiency (%)

60 60

40 40

20 20

0 0
1 2 3 4 5 6 1 2 3 4 5 6 7

pH pH
100 100
Y Y
Fe Fe
c) Sc d) Sc
80 Nd 80 Nd
Al Al
Precipitation Efficiency (%)
Precipitation Efficiency (%)

60 60

40 40

20 20

0 0
1 2 3 4 5 6 7 1 2 3 4 5 6 7

pH pH

Figure 3. Precipitation behavior of Fe(III), Al, Sc, Nd, and Y in HCl media with the addition of (a)
Figure 3. Precipitation behavior of Fe(III), Al, Sc, Nd, and Y in HCl media with the addition of
CaCO3; (b) NaOH; (c) KOH; (d) NH4OH.
(a) CaCO3 ; (b) NaOH; (c) KOH; (d) NH4 OH.
Table 5. (a) Critical pH values for 70% Fe(III) removal from the system by hydroxide precipitation
5. (a)
Tableand Critical pH
precipitation values
% of for 70% Fe(III)
the constituent removal
ions with fromofthe
selectivity Fe system
over Sc;by
(b)hydroxide
Critical pHprecipitation
values for and
precipitation % of the constituent ions with selectivity of Fe over Sc; (b) Critical
above 90% Fe(III) removal from the system by hydroxide precipitation and precipitation pH values for%above
of the constituent
90% Fe(III) removal fromionsthewith selectivity
system of Fe over
by hydroxide Sc.
precipitation and precipitation % of the constituent
ions with selectivity of Fe over Sc.
(a)
Precipitation Agent pH Fe (%) Sc (%) Al (%) Y (%) Nd (%) SFe/Sc
Limestone 2.81 70.7 (a)
25.2 15.3 15.1 15.6 7
Precipitation Agent
NaOH pH Fe (%)70.3 Sc (%)
2.97 14.4 Al9.9
(%) Y (%) 10.6Nd (%)14
5.8 SFe/Sc
Limestone KOH 2.81 3.0570.7 75.1 25.216.7 8.2
15.3 10.6
15.1 8.3 15.6 15 7
NaOH NH4OH 2.97 2.8270.3 71.3 14.4 9.6 4.4
9.9 1.05.8 0.2 10.6 24 14
KOH 3.05 75.1 16.7
(b) 8.2 10.6 8.3 15
NH4 OH 2.82 71.3 9.6 4.4 1.0 0.2 24
Precipitation Agent pH Fe (%) Sc (%) Al (%) Y (%) Nd (%) SFe/Sc
Limestone 2.95 91.2 (b)
28.6 25.9 29.4 24.2 25
NaOH pH
Precipitation Agent 3.10
Fe (%)91.4 Sc (%)
18.9 11.1
Al (%) 6.9
Y (%) 12.3Nd (%)46 SFe/Sc
Limestone KOH 2.95 3.2291.2 96.3 28.621.0 9.1
25.9 11.1
29.4 11.5 24.2 98 25
NaOH NH4OH 3.10 2.9591.4 93.8 18.913.2 5.2
11.1 1.16.9 0.3 12.3 100 46
KOH 3.22 96.3 21.0 9.1 11.1 11.5 98
NH4 OH 2.95 93.8 13.2 5.2 1.1 0.3 100
Metals 2018, 8, 314 8 of 13

3.2.2.Metals 2018, 8, x FOR PEER REVIEW

Phosphate Precipitation 8 of 13

3.2.2.
Figure Phosphate
4 showsPrecipitation
the precipitation behavior with the addition of different phosphate donors into a
syntheticFigure
solution with
4 shows theHClprecipitation
media. Thebehavior
selective precipitations
with the addition ofof different
scandium and irondonors
phosphate were triggered
into
immediately
a syntheticwith the addition
solution of a dibasic
with HCl media. phosphate
The selective solution,ofand
precipitations 60% precipitation
scandium and iron were efficiency
triggered was
attained, with below
immediately 10%
with the co-precipitation
addition of other elements
of a dibasic phosphate at pH
solution, and 60% values around
precipitation 1.5 in was
efficiency all cases.
Afterattained, with below
that point, further10% co-precipitation
addition of other ions
of phosphate elements at pH values
resulted in thearound
growth 1.5of
in both
all cases. After and
targeted
that point,elements.
non-targeted further addition
The best of phosphate
selectivityionswithresulted
90% ofinScthe growth of both
precipitation targetedwas
efficiency andreached
non- at
targeted elements. The best selectivity with 90% of Sc precipitation efficiency
pH 2.2 with the addition of K2 HPO4 , with co-precipitation yields of 27%, 26%, and 43% for Al, was reached at pH 2.2Y, and
with the addition of K2HPO4, with co-precipitation yields of 27%, 26%, and 43% for Al, Y, and Nd,
Nd, respectively. (NH4 )2 HPO4 showed a similar performance with slightly increased co-precipitation
respectively. (NH4)2HPO4 showed a similar performance with slightly increased co-precipitation
levels corresponding to 36% for Al, 23% for Y, and 42% for Nd when Sc precipitation hit 90%. Critical
levels corresponding to 36% for Al, 23% for Y, and 42% for Nd when Sc precipitation hit 90%. Critical
pH values for above
pH values 90%90%
for above Sc recovery
Sc recovery from
fromthethe
solution
solutionsystem
systemby bydibasic
dibasic phosphate precipitation and
phosphate precipitation
the precipitation efficiencies
and the precipitation of the of
efficiencies constituent ionsions
the constituent can can
be found
be found in in
Table
Table6.6.
Since introducing
Since introducing newnew ions
ionsinto
intothe
the system
system is isunwanted
unwanted throughout
throughout the the precipitation
precipitation process,
process,
(NH4(NH
)2 HPO
4)2HPO
4 was4 selected
was selected asasthe
the best
best precipitation
precipitation agent
agent for
for selective
selective precipitation
precipitation of Sc of Sc
from from bauxite
bauxite
residue leachates.
residue leachates.
100 100
Y Y
a) Fe b) Fe
Sc Sc
80 Nd 80 Nd
Al Al
Precipitation Efficiency (%)
Precipitation Efficiency (%)

60 60

40 40

20 20

0 0
1 2 3 1 2 3

pH pH
100
Y
c) Fe
Sc
80 Nd
Al
Precipitation Efficiency (%)

60

40

20

0
1 2 3

pH

Figure 4. Precipitation of the elements in HCl media with the addition of (a) Na2HPO4; (b) K2HPO4
Figure 4. Precipitation of the elements in HCl media with the addition of (a) Na2 HPO4 ; (b) K2 HPO4
and (c) (NH4)2HPO4.
and (c) (NH4 )2 HPO4.
Table 6. Critical pH values for above 90% Sc recovery from the solution system by dibasic phosphate
6. Critical pH
Tableprecipitation values for above 90% Sc recovery from the solution system by dibasic phosphate
and precipitation efficiencies of the constituent ions.
precipitation and precipitation efficiencies of the constituent ions.
Precipitation Agent pH Fe(III) (%) Sc (%) Al (%) Y (%) Nd (%)
PrecipitationNa 2HPO4
Agent pH 2.32 Fe(III)
92.1
(%) 90.9
Sc (%) 39.2
Al (%) 26.0 Y (%)47.3 Nd (%)
K2HPO4 2.19 90.1 90.2 27.5 26.3 42.7
Na2 HPO 4 4)2HPO4 2.32
(NH 2.27 92.1
90.4 90.9
90.3 36.339.2 22.9 26.042.1 47.3
K2 HPO4 2.19 90.1 90.2 27.5 26.3 42.7
(NH4 )2 HPO4 2.27 90.4 90.3 36.3 22.9 42.1
3.3. Successive Precipitation in All Media
TakingPrecipitation
3.3. Successive all the findings intoMedia
in All consideration, triple-staged precipitation processes for the synthesis
of a scandium concentrate from bauxite residue solutions can be proposed using the described
Taking all the findings into consideration, triple-staged precipitation processes for the synthesis
of a scandium concentrate from bauxite residue solutions can be proposed using the described mineral
Metals 2018, 8, 314 9 of 13

acid systems. The proposed precipitation processes are summarized in Figure 5, and the variation
in concentration with the successive addition of the precipitation agents in H2 SO4 , HNO3 , and HCl
media can be found in Figure 5a–c, respectively. Three precipitation regions labelled as regions 1, 2,
and 3 are shown in all graphs. Region 1 is the removal of Fe from a solution with minimum Sc loss by
the addition of NH4 OH until a specified pH is reached for each media. The residue obtained from this
step, which was enriched in Fe, was filtered and removed from the solution. The critical pH ranges
for the first Fe removal step were established as 3.3–3.4 for H2 SO4 and 2.8–2.9 for HNO3 and HCl.
Region 2 denotes the second Fe removal step by further addition of NH4 OH that promotes higher Fe
precipitation with a low amount of Sc co-precipitation. The precipitate was then filtered and could
be recycled into the initial feed, which minimized Sc losses during the second Fe removal step and
provided a seeding agent for better Fe separation. The pH ranges for this step were between 3.6–3.7
for H2 SO4 , 3.1–3.2 for HNO3 , and 3.0–3.1 for HCl media.
After removing more than 95% of Fe using NH4 OH with low Sc loss, a phosphate precipitation
step was applied in the third part of this successive precipitation process. It was observed that the
phosphate precipitation with dibasic phosphates showed similar performances regardless of the
system. The pH of all systems was adjusted to 2.0 in advance, with the purpose to avoid unwanted
hydroxide precipitation which can be triggered at pH levels above 3.0. Dibasic phosphate solutions
were added into all systems until reaching a pH range between 2.5 and 2.6, where the best selectivity
for Sc was reached. Since Fe was removed in previous steps, both the amount of precipitant solution
and the co-precipitation levels of the constituent ions were decreased. For all media, more than 95% of
Sc and Fe was recovered, while approximately 15% of Al and 10% of Y and Nd were co-precipitated
during the process. The compositions of the resulted concentrate in the form of mixed phosphates can
be found in Table 7.

Table 7. Compositions of each concentrate obtained after their successive selective precipitation
as phosphates.

H2 SO4 HNO3 HCl

ScPO4 (%) 66 56 51
FePO4 (%) 13 12 13
AlPO4 (%) 19 23 26
YPO4 (%) 1 3 5
NdPO4 (%) 1 6 5

In all cases, Sc concentrates containing more than 50% ScPO4 were synthesized. Since the
precipitation patterns of the ions were more discrete in the sulfuric acid media, the best results were
obtained in this media, with 66% ScPO4 . The major impurity element was found to be Al in the
concentrate, which can be easily processed and removed with basic purification operations.
Metals 2018, 8, 314 10 of 13
Metals 2018, 8, x FOR PEER REVIEW 10 of 13

2.00 2.25 2.50

450

a)
375
Fe
Al
300 Sc

Concentration (mg/L)
Nd
Y
1 2 3
225

150

75

0
1.3 1.7 2.1 2.5 2.9 3.3 3.6
pH
2.00 2.25 2.50
375

b)
300
Fe
Al
Sc
Concentration (mg/L)

Nd
225 Y
1 2 3

150

75

0
1.3 1.8 2.3 2.8 3.1
pH
2.00 2.25 2.50
375

c)
300

Fe
Concentration (mg/L)

Al
225 Sc
Nd
Y
1 2 3

150

75

0
1.3 1.8 2.3 2.8 3.0
pH

5. Triple-staged
FigureFigure successive
5. Triple-staged precipitation
successive with
precipitation withNH
NH4 OH
4OH and
and (NH
(NH44))22HPO
HPO4 4from
from(a)(a)
H2H2 SO
SO 4;
4; (b)

(b) HNO
HNO and (c) HCl media.
3 3 and (c) HCl media.
Metals
Metals 2018,
2018, 8,
8, x314
FOR PEER REVIEW 1111ofof 13
13

4. Assessment and Conclusions

4. Assessment and Conclusions
A triple-step precipitation route is proposed to refine a scandium concentrate from synthetic
bauxiteA triple-step precipitation
residue leachates, and the route is proposed
effect in aqueoustomedia refine was
a scandium concentrate
investigated. from synthetic
In all media, NH 4OH
bauxite residue leachates, and the effect in aqueous media was investigated.
showed better selectivity and performance than the other hydroxides, similar to the results obtained In all media, NH4 OH
showed
in H2SO4better
media. selectivity
While HNO and3performance
and HCl media thanshowed
the other hydroxides,
quite similar to
similar patterns the results
during obtained
precipitation,
in H2differed
they SO4 media. withWhile
respectHNO 3 and
to the H2SO HCl mediaespecially
4 media, showed quite in thesimilar
pH rangespatterns during precipitation,
of precipitation, because
they differed with respect to
of the nature of the formed complexes. the H 2 SO media, especially in the pH ranges of precipitation,
4 Although Sc loss during the Fe removal step in HCl media because of
the nature
was observed of the
to beformed complexes.
relatively AlthoughtoScthe
higher compared loss during
other casesthe
asFe removal
a result step
of the in HCl media
interaction betweenwas
observed
Fe 3+ and Cl to -be relatively
ions, higher
still the comparedcontained
concentrate to the other casesthan
more as a 50%
resultofofSc.
the The
interaction
proposed between Fe3+
selective
and Cl− ions,route
precipitation still the concentrate
to produce ScPOcontained
4 concentratemore can
thanbe50%seenof in
Sc.Figure
The proposed
6. selective precipitation
routeAtoscandium
produce concentrate
ScPO4 concentrate
containing can bemore seen in Figure
than 50% of6. ScPO4 was successfully obtained with this
A scandium concentrate containing more
selective precipitation route from synthetic solutions. Depending than 50% of ScPO4 wason thesuccessfully obtained
impurity level, with this
the aqueous
selective
media, precipitation
and the initial route from synthetic
composition of the feedsolutions. Depending
solution, on theofimpurity
the amount ScPO 4 inlevel, the aqueous
the synthesized
media, andcould
concentrate the initial composition
vary between 15% of and the65%.feed solution,
Since the amount
the concentrate of ScPOmore
contained 4 in the synthesized
scandium than
concentrate could vary between 15% and 65%. Since the concentrate
the other elements, the processing and the purification of this concentrate will be much easier contained more scandium than
to
the otherThe
perform. elements,
removal theof processing and the purification
the most problematic element for ofscandium
this concentrate will bewhich
precipitation, muchiseasier
iron, tois
perform.
the main The
obstacleremoval of the mostprocesses
for advanced problematic element
to obtain a for scandiumscandium
high-purity precipitation, whichFurthermore,
product. is iron, is the
main
the obstacle for
remaining advanced
solution after processes
precipitation to obtain
can bea treated
high-purity scandium
for further REEs product.
recovery Furthermore,
through other the
remaining solution after precipitation can be treated for further REEs recovery through other means.
means.

Figure
Figure 6.
6.Proposed
Proposedselective
selectiveprecipitation
precipitation route
route to
to synthesize
synthesize aa ScPO
ScPO44concentrate
concentrateusing
usingthree
threemineral
mineral
acid media.
acid media.

Author
Author Contributions:
Contributions: B.Y.,
B.Y., C.D.,
C.D., and
and B.F.
B.F. conceived
conceived and
and designed
designed the
the experiments;
experiments; B.Y.
B.Y. performed
performed the
the
experiments; B.Y.
experiments; B.Y. analyzed the data; B.Y. wrote the paper; B.Y., C.D., and B.F. reviewed the paper.
wrote the paper; B.Y., C.D., and B.F. reviewed the paper.
Funding: This
Funding: This research
researchwas funded
was by European
funded Community’s
by European Horizon
Community’s 2020 Programme
Horizon under grant
2020 Programme agreement
under grant
number 636876.
agreement number 636876.
Acknowledgments: The research leading to these results has received funding from the European Community’s
Acknowledgments: The research leading to these results has received funding from the European Community’s
Horizon 2020 Programme ([H2020/2014–2019]) under Grant Agreement no. 636876 (MSCA-ETN REDMUD).
Horizon 2020 Programme
This publication ([H2020/2014–2019])
reflects only the author’s view,under Grantthe
exempting Agreement no.from
Community 636876
any(MSCA-ETN REDMUD).
liability. Project website:
This publication reflects only the
http://www.etn.redmud.org. Theauthor’s view, Wenzhong
authors thank exempting Zhang
the Community from
and Dzenita any liability.
Avdibegovic Project
for their website:
supports in
ICP Measurements.
http://www.etn.redmud.org. The authors thank Wenzhong Zhang and Dzenita Avdibegovic for their supports
in ICP Measurements.
Conflicts of Interest: The authors declare no conflict of interest.
Conflicts of Interest: The authors declare no conflict of interest.
References
References
1. UNFCCC. Adoption of the Paris Agreement. Report No. FCCC/CP/2015/L.9/Rev.1. Available online:
1. http://unfccc.int/resource/docs/2015/cop21/eng/l09r01.pdf
UNFCCC. (accessed on 15 January 2018).
Adoption of the Paris Agreement; Report No. FCCC/CP/2015/L.9/Rev.1. Available online:
2. Røyset, J.; Ryum, N. Scandium in aluminium alloys. Int. Mater.
http://unfccc.int/resource/docs/2015/cop21/eng/l09r01.pdf Rev. on
(accessed 15 50,
2005, 19–44. [CrossRef]
January2018).
3.
2. Gambogi,
Røyset, J. USGSN.Minerals
J.; Ryum, Scandium Information: Scandium;
in aluminium U.S.
alloys. Int.Geological
Mater. Rev.Survey: Reston,
2005, 50, 19–44.VA, USA; pp. 146–147.
4.
3. Lathabai, S.;
Gambogi, Lloyd,Minerals
J. USGS P. The effect of scandium
Information: on the
Scandium; microstructure,
U.S. mechanical
Geological Survey: properties
Reston, VA, USA;and
pp.weldability
146–147.
4. of a cast Al–Mg
Lathabai, alloy.
S.; Lloyd, Actaeffect
P. The Mater.of2002, 50, 4275–4292.
scandium [CrossRef] mechanical properties and weldability
on the microstructure,
of a cast Al–Mg alloy. Acta Mater. 2002, 50, 4275–4292.
Metals 2018, 8, 314 12 of 13

5. Marquis, E.; Seidman, D. Nanoscale structural evolution of Al3 Sc precipitates in Al-Sc alloys. Acta Mater.
2001, 49, 1909–1919. [CrossRef]
6. Yamamoto, O. Solid oxide fuel cells: Fundamental aspects and prospects. Electrochim. Acta 2000, 45,
2423–2435. [CrossRef]
7. European Commission. Study on the Review of the List of Critical Raw Materials: Executive Summary;
Directorate-General for Internal Market, Industry, Entrepreneurship and SMEs: Brussels, Belgium, 2017;
pp. 1–93.
8. Power, G.; Gräfe, M.; Klauber, C. Bauxite residue issues: I. Current management, disposal and storage
practices. Hydrometallurgy 2011, 108, 33–45. [CrossRef]
9. Evans, K. The history, challenges, and new developments in the management and use of bauxite residue.
J. Sustain. Metall. 2016, 2, 316–331. [CrossRef]
10. Li, G.; Liu, M.; Rao, M.; Jiang, T.; Zhuang, J.; Zhang, Y. Stepwise extraction of valuable components from
red mud based on reductive roasting with sodium salts. J. Hazard. Mater. 2014, 280, 774–780. [CrossRef]
[PubMed]
11. Liu, Y.; Naidu, R. Hidden values in bauxite residue (red mud): Recovery of metals. Waste Manag. 2014, 34,
2662–2673. [CrossRef] [PubMed]
12. Paramguru, R.; Rath, P.; Misra, V. Trends in red mud utilization—A review. Miner. Process. Extr. Metall. Rev.
2004, 26, 1–29. [CrossRef]
13. Alkan, G.; Yagmurlu, B.; Cakmakoglu, S.; Hertel, T.; Kaya, Ş.; Gronen, L.; Stopic, S.; Friedrich, B. Novel
approach for enhanced scandium and titanium leaching efficiency from bauxite residue with suppressed
silica gel formation. Sci. Rep. 2018, 8, 5676. [CrossRef] [PubMed]
14. Avdibegović, D.; Regadío, M.; Binnemans, K. Recovery of scandium(III) from diluted aqueous solutions by a
supported ionic liquid phase (silp). RSC Adv. 2017, 7, 49664–49674. [CrossRef]
15. Narayanan, R.P.; Kazantzis, N.K.; Emmert, M.H. Selective process steps for the recovery of scandium from
jamaican bauxite residue (red mud). ACS Sustain. Chem. Eng. 2018, 6, 1478–1488. [CrossRef]
16. Ochsenkühn-Petropulu, M.; Lyberopulu, T.; Parissakis, G. Selective separation and determination of
scandium from yttrium and lanthanides in red mud by a combined ion exchange/solvent extraction method.
Anal. Chim. Acta 1995, 315, 231–237. [CrossRef]
17. Onghena, B.; Borra, C.R.; Van Gerven, T.; Binnemans, K. Recovery of scandium from sulfation-roasted
leachates of bauxite residue by solvent extraction with the ionic liquid betainium bis (trifluoromethylsulfonyl)
imide. Sep. Purif. Technol. 2017, 176, 208–219. [CrossRef]
18. Wang, W.; Cheng, C.Y. Separation and purification of scandium by solvent extraction and related technologies:
A review. J. Chem. Technol. Biotechnol. 2011, 86, 1237–1246. [CrossRef]
19. Wang, W.; Pranolo, Y.; Cheng, C.Y. Recovery of scandium from synthetic red mud leach solutions by solvent
extraction with D2HPA. Sep. Purif. Technol. 2013, 108, 96–102. [CrossRef]
20. Zhang, W.; Koivula, R.; Wiikinkoski, E.; Xu, J.; Hietala, S.; Lehto, J.; Harjula, R. Efficient and selective recovery
of trace scandium by inorganic titanium phosphate ion-exchangers from leachates of waste bauxite residue.
ACS Sustain. Chem. Eng. 2017, 5, 3103–3114. [CrossRef]
21. Yagmurlu, B.; Alkan, G.; Xakalashe, B.; Friedrich, B.; Stopic, S.; Dittrich, C. Combined saf smelting and
hydrometallurgical treatment of bauxite residue for enhanced valuable metal recovery. In Proceedings of the
35th International Conference and Exhibition of ICSOBA, Hamburg, Germany, 2–5 October 2017.
22. Yagmurlu, B.; Dittrich, C.; Friedrich, B. Precipitation trends of scandium in synthetic red mud solutions with
different precipitation agents. J. Sustain. Metall. 2017, 3, 90–98. [CrossRef]
23. Alkan, G.; Xakalashe, B.; Yagmurlu, B.; Kaussen, F.; Friedrich, B. Conditioning of red mud for subsequent
titanium and scandium recovery–a conceptual design study. World Metall. ERZMETALL 2017, 70, 5–12.
24. Horovitz, C.T. Scandium its Occurrence, Chemistry Physics, Metallurgy, Biology and Technology; Elsevier:
New York, NY, USA, 2012.
25. Vickery, R.C. The Chemistry of Yttrium and Scandium; Oxford: Oxford, UK, 1960; Volume 2.
26. Ahrland, S. How to distinguish between inner and outer sphere complexes in aqueous solution.
Thermodynamic and other criteria. Coord. Chem. Rev. 1972, 8, 21–29. [CrossRef]
27. Spiro, T.G.; Revesz, A.; Lee, J. Volume changes in ion association reactions. Inner-and outer-sphere complexes.
J. Am. Chem. Soc. 1968, 90, 4000–4006. [CrossRef]
Metals 2018, 8, 314 13 of 13

28. Lucas, S.; Champion, E.; Bregiroux, D.; Bernache-Assollant, D.; Audubert, F. Rare earth phosphate powders
REPO4 ·NH2 O (Re = La, Ce or Y)—Part i. Synthesis and characterization. J. Solid State Chem. 2004, 177,
1302–1311. [CrossRef]
29. Beltrami, D.; Deblonde, G.J.-P.; Bélair, S.; Weigel, V. Recovery of yttrium and lanthanides from sulfate
solutions with high concentration of iron and low rare earth content. Hydrometallurgy 2015, 157, 356–362.
[CrossRef]

© 2018 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/).