Separation and Purification Technology 256 (2021) 117770

Contents lists available at ScienceDirect

Separation and Purification Technology

journal homepage: www.elsevier.com/locate/seppur

Extraction and optimization of neodymium from molten

fluoride electrolysis
Prince Sarfo, Avimanyu Das, Courtney Young *
Metallurgical and Materials Engineering, Montana Tech, 1300 West Park Street, Butte, MT 59701, United States

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

Keywords: Most rare earth elements (REEs) are critical materials of extreme prominence. Because of their criticality, their
Electrolysis convalescence into metal from rare earth oxide (REO) and their recycling from end-of-life materials are essential
Statistical modeling for the proficient use and safeguard of natural resources. With pyrometallurgical approaches using potential
Neodymium
versus partial oxidation pressure (E-pO2-) diagrams, the conditions for making neodymium are established by
Molten fluoride
Optimization
which REOs are dissolved in a molten halide bath to form pure Nd metal in the electrolysis process. This paper
also expands on how the diagrams are used and verified through design of experiment studies, which result in
statistical rejoin models based on molten salt electrolysis for extracting neodymium from its oxide in a fluoride
bath.

1. Introduction
Nd2O3 + 2F- = 2NdOF + O2– (2)
The production of powerful magnets and the need to fabricate more -
NdOF + 4F = [NdOF5] 4-
(3)
useful high-tech products have increased the demand of neodymium
(Nd) and other rare earth elements (REEs). Most of these materials are
4- +
[NdOF5] = NdO + 5O + 5/2F2 + 5e 2- -
(4)
obtained through mining and recycling [1]. REE recovery processes
NdO+ + 3e- = Nd + O2– (5)
from both mining and recycling are based on hydrometallurgy, pyro­
metallurgy or a combination thereof. In order to separate compounds of 2– 2–
O or oxygen (O2) generated from O in the process will react with
akin chemical and physical properties, hydrometallurgical approaches graphite (C) to form carbon monoxide (CO) and carbon dioxide (CO2)
of extraction for individual rare earth oxides (REOs) from their con­ [10]. Also, in the [NdF6]3- formation, NdF3 reacts with excess F-, which
centrates have proven to be advantageous and both metallothermic can be shown to co-exist in the Nd3+ region in Fig. 1. The [NdF6]3-
reduction and molten salt electrolysis are currently the most commonly further reacts to produce Nd at the cathode according to reactions (6)–
used pyrometallurgical methods in this regard [2–5]. (8) [6,7]:
As EH-pH diagrams are used for aqueous systems, so are E-pO2- di­
agrams for molten salts; but the choice of reference potential is depen­ NdF3 + 3F- = [NdF6]3- (6)
dent on the content of the molten bath. However, for aqueous systems, 3-
[NdF6] = Nd 3+
+ 3F2 + 6e -
(7)
the electrolytic bath was water making the Standard Hydrogen Elec­
trode (SHE) its universal choice. For Nd electrolysis in a fluoride bath, Nd3+ + 3e- = Nd (8)
Nd2O3 dissolves in the presence of NdF3 and excess F- to form oxy­
Likewise, the dissolution of a solute in a solvent is important when it
fluorides as shown in Fig. 1 [6–8]. The formation of [NdOF5]4- also takes
comes to electrochemical extraction. Because, as the solute (Nd2O3) is
place when the NdOF reacts with excess F- in the melt, which reacts
consumed and the product (Nd metal) is formed at the cathodic elec­
further to form Nd at the cathode according to reactions (1)–(5) [9–11]:
trode, concentration gradients develop around the electrode up to the
1/3Nd2O3 + 1/3NdF3 = NdOF (1) bulk of the solution, meaning that reactants move towards the electrode
surface, metal is deposited, and by-products leave the interfacial region.
Or

* Corresponding author.
E-mail address: cyoung@mtech.edu (C. Young).

https://doi.org/10.1016/j.seppur.2020.117770
Received 22 June 2020; Received in revised form 7 September 2020; Accepted 17 September 2020
Available online 23 September 2020
1383-5866/© 2020 Elsevier B.V. All rights reserved.
P. Sarfo et al.
Fig. 1. E-pO2- Diagram of Nd in NdF3-CaF2-LiF Melt with C as a Compo­
nent [9].
Fick’s law is therefore applicable [12,13].
This paper presents an in-depth study of molten salt electrolysis to
recover Nd from its oxide in a fluoride bath utilizing the E-pO2- diagram
in Fig. 1 as an aid. A detailed statistical analysis is also performed to
develop a rejoin model with temperature, oxide addition and dwelling
time (the time the oxide remains in the bath at the target temperature
before electrowinning) as the condition variables in order to optimize
the process. To the best of our knowledge, the statistical analysis and
optimization of this system has not been attempted before. This con­
stitutes the novelty of the present work. The applicable diagram from
that study is reproduced in Fig. 1 to help predict conditions and eluci­
date reactions for making Nd metal as well as other species that could
exist when Nd2O3 is dissolved in a CaF2/LiF/NdF3 molten salt. Oper­
ating conditions that were varied including the amount of Nd2O3 added
are length of time the melt was left to sit at the target temperature (i.e.,
dwelling time), and bath temperature as discussed later.
2. Experimental
Initial materials and products from this study were characterized
using X-ray Diffractometry (XRD), Scanning Electron Microscopy with
Energy Dispersive X-ray (SEM-EDX), Inductively Coupled Plasma – Mass
Spectrometry (ICP-MS), and Inductively Coupled Plasma – Optical
Emission Spectrometry (ICP-OES). These different experimental and
methodology are discussed in the ensuing sections following a brief
description of the materials and procedures used in the statistically
designed experiments.
2.1. Materials
Reagents used for this work were of analytical grade. Neodymium
fluoride was purchased from Stem Chemicals, calcium fluoride and
neodymium oxide were obtained from Beantown Chemical, molybde­
num wire was procured from Goodfellow Cambridge Ltd., and carbon
rod as well as graphite crucibles were acquired from National Carbon
Co., Inc. Ultra-high purity (UHP) argon gas was attained from General
Distributing Company.
2.2. Methodology
Molten salt electrolytic experiments were conducted using a bath
composition of 72 mol% LiF, 18 mol% CaF2 and 10 mol% NdF3 at high
2
Separation and Purification Technology 256 (2021) 117770
chemical purity due to the use of analytical grade materials. This
composition was used because LiF has a large electrochemical window
and the addition of CaF2 as a flux for temperature and viscosity reduc­
tion. However, CaF2 makes the electrochemical window more negative
for lower activity of Li+ [14]. These materials were combined and
thoroughly homogenized using a mortar and pestle. After placing the
mixture in a graphite crucible, it was heated with a Thermo Scientific
Lindberg Blue-M furnace at a rate of 6◦ C per minute. Throughout each
experiment, argon gas was used to purge the system and prevent
contamination from outside gases. For the initial experiment the con­
ditions were: 1050◦ C temperature, addition of 2wt% Nd2O3 and 3 hours
of dwelling time for the oxide in the bath before electrowinning. For the
optimization of the process using statistical design, the following ranges
were used for the condition variables: temperature (950–1050 ◦ C),
Nd2O3 addition (2–10wt%), and dwelling time for the oxide in the bath
before electrowinning (1–5 hours). When the temperature was reached,
the molten mixture was left to dwell before polarizing. Carbon rod with
an average diameter of 6.16 mm was used as the anode and molybde­
num rod with a diameter of 1 mm was employed as the cathode. Based
on the reading from Fig. 1, a voltage of 4.9 V was applied using a
Lambda model LK 340-A power supply with a mean current of 2.5A to
record from the reading gauge. Initially, the current fluctuated within
4% (2.5± 0.1 A) and when it stabilized, the fluctuation was less than 1%.
When the reduction time was reached (which changed due to the
composition of the mixture, but the most metal recovery was obtained in
1 hour) all experiments stopped on their own and molten baths were
allowed to cool to room temperature. After removing the crucible as­
sembly from the furnace, solidified products were carefully removed
from the crucible so that they could be analyzed. The crucible and
electrodes were cleaned and reused in subsequent experiments.
2.3. Characterization using XRD
X-ray diffraction was carried out with a Rigaku Ultima IV X-ray
Diffractometer (XRD) using Cu-Kα radiation at 40 kV and 40 mA. This
was used to quantify the materials and determine the various phases of
the products following the experiments upon solidification and cooling.
2.4. SEM/EDX Studies
A Scanning Electron Microscope coupled with an Energy Dispersive
X-ray (SEM-EDX) analyzer was employed to determine the chemical
compositions of all phases in the metal and the spent bath. The SEM/
EDX system was a TESCAN TIMA with a tungsten filament and an EDAX
Z2 analyzer. Samples were hand-separated, homogenized as needed,
and cold-mounted in epoxy using molds approximately 25 mm in
diameter and 10 mm in thickness. Resulting mounts were ground and
polished to a smooth finish and then conductively-coated with gold to
obtain SEM images by Back-Scattered Electron (BSE) detection and EDX
analyses helped determine the chemical compositions of all products.
2.5. ICP Analysis
Aqua Regia (3:1 v/v HCl to HNO3) was used to digest the metal.
Resulting solutions were analyzed by ICP-MS and ICP-OES to determine
the elemental content of the resulting metal.
3. Results and discussions
3.1. XRD Findings
XRD patterns in Fig. 2a-e confirm the crystalline phases of samples
used. Fig. 2e shows a typical example of spent bath after electrolysis and
cooling. Each of the three fluorides are observed along with neodymium
oxyfluoride (NdOF) when the Nd2O3 were added (Fig. 2e). The NdOF is a
secondary phase formed from Nd2O3 reacting with NdF3 and excess F-
P. Sarfo et al.
Fig. 2. XRD patterns of feed materials and spent bath after solidification.
similar to Reaction (1) and (2). [15–17]
3.2. Electrolysis results
Fig. 3a and b illustrate example products of Nd metal and spent bath
obtained for one experiment. The spent bath shows two layers. The
lower layer is made of partially dissolved oxide (NdOF) for 2wt% Nd2O3.
This material settled to the bottom of the crucible because it is quite
dense. By comparison, the upper layer is light unspent electrolyte.
Essentially, the density of the oxide is higher than that of the salt.
3.3. SEM/EDX observations
The BSE image in Fig. 4 shows there are three distinct phases in the
spent bath from 2wt% Nd2O3 addition. EDX analysis of the three indi­
cated points in Fig. 4a-c shows each has a distinctive chemistry: white
areas (a) appear to be NdOF (neodymium oxyfluoride) correlating to the
heavy, undissolved oxide illustrated in Fig. 2e and 3b; black areas (b) are
Fig. 3. (a) metal and (b) solidified spent bath recovered from the electrolysis process of 2wt% Nd2O3 addition.
3
Separation and Purification Technology 256 (2021) 117770
attributed to LiF; and gray areas (c) are speculated to contain the three
fluoride salts, which identifies as light, unspent electrolyte discussed
about Fig. 3. It is important to note that Li was not observed in Fig. 3b-c
because its intensity is too low to be detected.
Similarly, the BSE image and EDX analysis of the recovered metal is
shown in Fig. 5. Results clearly indicate that the micrograph exemplifies
pure Nd metal. It is possible that Li metal could be present because, as
noted, it is not measurable with this method. To ensure that Li is not
present and to verify that the proper voltage was applied in the exper­
iments, additional analytical techniques had to be used.
From both the XRD and SEM/EDX analysis presented, it can be
concluded that NdOF forms in the extraction process when the oxide
reacts with either NdF3 or F- at high temperature and likely precipitates
out during solidification and cooling. This rare earth oxyfluoride is an
insoluble product that can have little to no participation in the elec­
trolytic reactions because it is not charged. This could be a factor that
decreases the current efficiency and therefore the amount of metal that
can be made [15].
P. Sarfo et al. Separation and Purification Technology 256 (2021) 117770

Element Weight % Atomic %

(b) (a) OK 4.32 18.81
(a) FK 11.00 40.31
NdL 84.68 40.88

(c)

Element Weight % Atomic %

(b) Element
FK
Weight %
100.00
Atomic %
100.00 (c) FK 43.32 71.38
CaK 28.94 22.60
NdL 27.75 6.02

Fig. 4. EDX analyses of the 2wt% Nd2O3 addition spent bath at the indicated spots on the BSE image.

Element Weight % Atomic %

NdL 100.00 100.00

Fig. 5. BSE image and EDX analysis of the metal obtained from 2wt% Nd2O3 addition.

3.4. ICP results
Table 1 shows the ICP-OES analysis results of the metal sample. The
purity level of the metal is indeed very high with traces of other metallic
elements, which are all contaminants from the materials used to perform
the experiment.
4. Response surface methodology
In order to moderate the number of experiments and capture the
interactions of the process variables in determining the process perfor­
mances, Response Surface Methodology (RSM) was pragmatic in the
analysis of the experiments. RSM is a mathematical and statistical
technique that employs empirical models to fit the experimental data
with reference to the detailed experimental design. With the process
responses not following a linear model [18,19] the Box-Behnken design
was employed for designing the experimental matrix to delineate the
response surfaces generated by the condition variables [20]. This design
opts for points in the experimental domain for a three-level factorial
arrangement in such a way that permits proficient approximation of the
first and second order coefficients for the mathematical model [21]. The
user decides a high level and a low level of each condition variable and
the mid-point is automatically identified for the point selection. A
number of experiments (usually 3 or 5) are conducted at the mid-point of
all variables to estimate the inherent variability associated with the
experimental technique.
For this work, the objectives were to maximize both metal weight
yield and Nd percent recovery. The reduction experiments were per­
formed per the RSM design of experiments developed using the statis­
tical software Design Expert 9 procured from Stat-Ease Inc., USA [22].
During the scoping tests, a number of condition variables were identified
that affected the metal yield and Nd recovery significantly. In order to

Table 1
Elemental Composition (%) of the Nd made by adding 2% Nd2O3 in a Fluoride Salt Electrolysis.
Sample composition Nd Ca Fe Li Mo Others

Wt (%) 99.992 0.00156 0.000915 0.004051 0.001312 4.1791E-06

4
P. Sarfo et al. Separation and Purification Technology 256 (2021) 117770

Table 2
DOE Outlining the Number of Experiments with Their Conditions and
Responses.
Run Nd2O3 Dwelling Temperature Metal Nd
No. (%) time (hrs) (◦ C) Yield (g) Recovery
(%)

1 6 5 950 0 0.00
2 6 1 950 0.0942 0.69
3 2 3 950 0 0.00
4 6 1 1050 1.8922 13.80
5 6 3 1000 1.0243 7.47
6 10 5 1000 1.3541 5.92
7 6 3 1000 1.3760 10.04
8 6 5 1050 2.3592 17.21
9 2 1 1000 1.1167 24.44
10 10 1 1000 2.0167 8.82
11 2 5 1000 0.9095 19.90
12 10 3 950 0.0833 0.36
13 6 3 1000 1.3047 9.52
14 6 3 1000 0.7787 5.71
15 6 3 1000 1.2439 9.07
16 10 3 1050 2.4939 10.91
17 2 3 1050 1.4845 32.49
Fig. 6. Shows the starting point of Nd2O3wt. % added to the bath.

limit the number of experiments, the three most important condition
variables were chosen, namely, the amount of Nd2O3 added, dwelling
time and bath temperature. These variables were also selected because
of their effects on solubility and, of course, the identified responses were
metal yield and Nd recovery per the objectives. The selected points in
with all condition variables being at midpoints. Results were analyzed to
develop statistically significant models for the responses of metal yield
and percentage of Nd recovery that are as shown in Equations (9) and
(10), in which A denotes Nd2O3 added, B denotes the dwelling time, and
C denotes temperature:

MetalYield(g) = 1.15 + 0.30A + 0.093B + 1.04C − 0.11AB + 0.23AC + 0.14BC + 0.066A2 + 0.14B2 − 0.20C2 − 0.31A2 B − 0.066A2 C (9)

NdRecovery(%) = 8.36 − 5.31A + 0.68B + 7.58C − 5.49AC + 1.03BC + 4.71A2 + 1.70B2 − 2.13C2 − 2.54A2 B + 3.18A2 C − 2.09AB2 (10)

the experimental domain and the responses obtained are discussed in

the following section along with the response surfaces, model equations
and the interactions of the condition variables. Finally, the optimization
process and point predictions are discussed.
With respect to the E-pO2- diagram and the electrolysis system, it was
assumed that the process will be free of oxygen. That is, all the Nd2O3
that is added will react with NdF3 to form NdOF and then Nd3+, which is
not the actual scenario. From Reaction (1), according to the design of
experiment, the amount of NdF3 is fixed at 10 mol% as part of the salt
composition so computation to determine the amount of oxygen in
Nd2O3 that will react with NdF3 to form NdOF was performed. For this,
0.51 mol% (3.15wt%) for the solubility of Nd2O3 in the bath was used
[23,15]. Using the formula pO2- = -log[O2–], the pO2- of the percentage
excess Nd2O3 values were calculated where X, Y and Z denotes the pO2-
for the percentage excess Nd2O3 addition for 2, 6, and 10wt%, respec­
tively, as indicated on the E-pO2- diagram shown in Fig. 6.
4.1. Effect of process conditions and optimization
It may be noted that, in Equations (9) and (10), the process variables
are all in terms of the coded values and lie in the range + 1 to − 1 with
the mid-point having a value of zero. Thus, for C, the maximum tem­
perature of 1050◦ C corresponds to + 1, the lowest temperature of 950 ◦ C
corresponds to − 1, and the midpoint temperature of 1000◦ C corre­
sponds to 0. In order to utilize these statistically significant equations,
the actual variable value needs to be converted to the coded form lying
between + 1 and − 1. For example, a temperature of 1010◦ C corresponds
to a value of + 0.2 for C. With the help of the model equations, the
system was optimized by maximizing the weight of metal yield and the
percentage of Nd recovery. The best solution was chosen and the opti­
mum condition along with the responses under these conditions for this
case are as follows:
Nd2O3 addition: 2.197wt% Metal Yield: 1.556 g

Temperature: 1050 ◦ C Nd Recovery: 35.48%

Table 2 shows the Design of Experiments (DOE) along with condition
variables and responses. In all, 17 experiments were performed in this
study with a temperature range of 950–1050◦ C, Nd2O3 addition range of
2–10wt%, and dwelling time range of 1–5 hours. Although current ef­
ficiency is important among other things, it was not considered in this
work but elsewhere [9]. Furthermore, 3 experiments were completed
Dwelling Time: 1 hr.
Using Equations (9) and (10), 3-D surface plots were generated to
illustrate the effects on the responses. Per Equation (9), the 3-D plots in
Fig. 7a-c were obtained to illustrate the dependence of metal yield on the

5
P. Sarfo et al.
process variables. The plots divulge the impact of temperature, Nd2O3
addition and dwelling time on metal yield. From Fig. 7b and c it can be
said that temperature throughout the investigation regime appears to
have more prominent effect on the process efficiency. With respect to
Nd2O3 addition, it can be observed from Fig. 7a and b that the metal
yield increases to some extent with Nd2O3 addition. Evidently, a higher
temperature (1050◦ C) is more conducive for a higher metal yield. With
very little dependence of yield on dwelling time, a short dwelling time is
preferable. On the other hand, Nd2O3 added had less significant impact
at low temperature but aided metal yield more significantly at higher
temperature.
As seen in Fig. 7a and b, metal yield increased slightly as the per­
centage of Nd2O3 added (2–10wt%) increased. For the 2wt% (which is
indicated as X in Fig. 6 as the starting point) it can be said that the Nd2O3
will react with the NdF3 to form NdOF. Some of the NdOF will then react
to form metallic Nd in accordance with Reactions (1), (3)–(5). Of course,
the same reaction will also occur for 6 and 10wt% Nd2O3 addition.
Again, from Fig. 6, it can be seen that the line between NdOF and Nd
begins to slope from pO2- value of 6.5 to 1. The distance between that
line and Li is the window at which electrolysis can occur. The window
became small upon moving towards pO2- value of 1 from 6.5. Because of
that, metal yield from 2wt% will be more than that obtained for 6 and
10wt% which agrees with the DOE values in Table 2.
Examination of Fig. 7a and c reveals that variation of the response
with dwelling time is very small throughout the operating regime, which
Fig. 7. a-c 3D plots of metal yield with respect to process variables at (a) 1050 ◦ C, (b) 1 h, and (c) 2.25% Nd2O3.
6
Separation and Purification Technology 256 (2021) 117770
implies that dwelling time does not have much of an impact on the metal
yield that results due to the excess F- in the melt for less oxide addition
and more NdOF in the higher oxide addition cases. When oxide is added
in the bath, as the reaction proceeds, F- will tend to increase and exceed
the amount of NdO+ and Nd3+ in the bath and thereby cause the reaction
to slow and even shut down.
Fig. 8a-c was generated from Equation (10) and illustrates the
response surfaces of Nd recovery in relations to process variables: Nd2O3
addition, dwelling time and temperature. From these plots, it is observed
that dwelling time again did not influence Nd recovery to any significant
extent. However, Nd2O3 addition, as observed in Fig. 8a and b, has
significant impact on Nd recovery. At high temperature, the recovery
dropped sharply with oxide addition. With increasing oxide addition,
the total Nd content of the melt increased but Nd recovery did not in­
crease proportionally. On the contrary, the recovery dropped and most
of the added Nd was essentially lost. Therefore, it is evident that a low
addition of Nd2O3 is better in terms of recovery of Nd. This is attributed
to all of the Nd2O3 reacting to form oxyfluoride per Reactions (1) and
(2). In view of its low solubility, only a small fraction of the oxyfluoride
would dissociate in excess F- into NdO+ and Nd3+ according to Reactions
(4) and (7) and the NdO+ and Nd3+ formed, would finally be converted
to metallic Nd by electrolysis per Reactions (5) and (8). At high tem­
perature, NdOF solubility is relatively high and therefore, the sharp drop
in Nd recovery with oxide addition is prominent. At low temperature,
the solubility is poor and then leads to low recovery with increasing
P. Sarfo et al.
Fig. 8. a-c 3D plots of Nd recovery with respect to process variables at (a) 1050 ◦ C, (b) 1 h, and (c) 2.25% Nd2O3.
oxide addition. According to Fig. 8b and c, temperature has a much
greater influence on Nd recovery and this impact was observed to
decrease slightly with increasing Nd2O3 addition. This is attributed to
higher oxide solubility because more NdOF forms at higher temperature.
Melt temperature was by far the most dominant factor on the process as
a whole.
With respect to temperature, it can be said that this study confirms Le
Chatelier’s Principle which predicts how much a system will respond to
a change in external conditions. Also, the solute dissolution rate is
affected by the addition of more solute (Nd2O3) to the solvent (molten
bath). But since Nd2O3 has a low solubility in most fluoride, it appears
that the addition of more Nd2O3 alters the Nd deposition mechanism
with respect to the bath to a limited extent as shown in Figs. 7 and 8
[24]. Dissolved oxide concentration cannot be increased too much by
adding excess oxide into the bath at constant temperature and fixed salt
composition. However, oxide addition can increase the kinetics of the
reactions since increasing the concentration of a reactant increases the
number of collisions between the reacting species per second and
therefore increases the reaction rate [25]. This will be true if all the
oxide dissolved. It does not happen due to the low solubility of the oxide
and addition of the oxide merely increases the Nd content of the melt by
means of forming oxyfluoride. The sharp drop in Nd recovery with
addition of oxide bears a testimony to the low solubility of oxide in the
fluoride bath. Of course, the solubility increases at elevated temperature
which is what is observed in the present investigation as the metal yield
and Nd recovery increased. But with more addition of oxide, the Nd
recovery dropped sharply due to the limited solubility even at higher
7
Separation and Purification Technology 256 (2021) 117770
temperature. Enhanced REO solubility leads to higher concentration of
Nd3+ cations in the melt. This is why the metal yield and Nd recovery
increases. But with increased Nd2O3 level and limited solubility even at
higher temperature the Nd recovery dropped sharply. The oxide addi­
tion merely increased the available Nd in the melt. Evidently, the sol­
ubility at any specified temperature is the limiting factor in this case.
Additionally, dissolution rates differ between different systems and,
because solute dissolution is also time dependent, how much a solute
dissolve in a solvent becomes an important factor. Substances with low
dissolution rates typically exhibit low solubility while substances with
high solubility lead to high dissolution rates.
Cogitating all the results discussed above, it is quite justified that a
high temperature with low Nd2O3 addition will help in the achievement
of better Nd metal recovery. Dwelling time does not appear to be so
critical beyond the one hour and therefore, a short dwelling time would
be advisable.
5. Conclusions
From this study, it is established that E-pO2- diagrams are used to
thermodynamically model the electrolysis of molten salts and the re­
covery of Nd metal is possible and metal produced using this technology
is much purer. In the casestudy for Nd electrolysis from LiF-CaF2-NdF3
baths, an average cathode potential of − 4.9 V would be needed to make
Nd under acidic condition.
From the analysis, it can be confirmed that not all of the Nd2O3
dissolves in the molten salt. Some reacts to form neodymium oxyfluoride
P. Sarfo et al.
(NdOF). As the Nd2O3 content increased (greater than 6wt%), the un­
dissolved Nd2O3 and NdOF show up in the spent bath. The NdOF and/or
the Nd2O3, being dense, settled at the bottom of the spent bath [25,26].
This is likely a factor that decreases the current efficiency which reduces
the amount of metal that can be made. This effect becomes worse at
excessively high oxide addition levels. Solubility of the oxide increases
at higher temperature but not very significantly. Thus, the metal re­
covery could be slightly more at high temperature and low oxide addi­
tion. But, with increased oxide addition, the recovery drops sharply even
at high temperature due to low solubility.
Using statistical design of experiments (DOE), process parameters
such as dwelling time, melt temperature and Nd2O3 addition were
defined in detail and a better process regime for the electrolysis process
was identified. It was established that higher temperature, low Nd2O3
addition, and short dwelling time will achieve the best performance in
Nd metal recovery and the metal yield from a fluoride bath by electro­
winning technique.
Declaration of Competing Interest
The authors whose names are listed on the manuscript certify that
this research was sponsored by the Army Research Laboratory and was
accomplished under Cooperative Agreement Number W911NF-15-2-
0020. The views and conclusions contained in this document are those of
the authors and should not be interpreted as representing the official
policies, either expressed or implied, of the Army Research Laboratory
or the U.S. Government. The U.S. Government is authorized to repro­
duce and distribute reprints for Government purposes notwithstanding
any copyright notation herein.
Acknowledgements
Research was sponsored by the Army Research Laboratory and was
accomplished under Cooperative Agreement Number W911NF-15-2-
0020. The views and conclusions contained in this document are those of
the authors and should not be interpreted as representing the official
policies, either expressed or implied, of the Army Research Laboratory
or the U.S. Government. The U.S. Government is authorized to repro­
duce and distribute reprints for Government purposes notwithstanding
any copyright notation herein. Gratefully acknowledged by the authors
is the scholarly advice received from Dr. John Murphy. Sincere thanks
are also extended to the Metallurgical & Materials Engineering
Department of MTU for the matching support to make this project
successful.
Separation and Purification Technology 256 (2021) 117770
References
[1] G. Charalampides, K. I.Vatalis, B. Apostoplos, & B. Ploutarch-Nikolas, (January 01,
2015). Rare Earth Elements: Industrial Applications and Economic Dependency of
Europe. Procedia Economics and Finance, 24, 126-135.
[2] M. Tanaka, T. Oki, K. Koyama, H. Narita, T. Oishi, 2013. Recycling of rare earths
from scrap. In: Bunzli, J.C.G., Pecharsky, V.K. (Eds.), Handbook on the Physics and
Chemistry of Rare Earths, vol. 43. Elsevier, Amsterdam, pp. 159e212. Chapter 255.
[3] Q. Tan, J. Li, X. Zeng, Critical Reviews in Environmental Science and Technology
45 (7) (2015) 749–776.
[4] A. Tuncuk, V. Stazi, A. Akcil, E.Y. Yazici, H. Deveci, Minerals Engineering 25 (1)
(2012) 28–37.
[5] B.J. Beaudry, and K.A. Gschneidner, (1978). Chapter 2 preparation and basic
properties of the rare earth metals. In: Gschneidner, K. A., B¨unzli, J. C., and
Pecharsky, V., eds. Handbook on the physics and chemistry of rare earths (pp.
173–232). Amsterdam: Elsevier.
[6] C. Huang, X. Liu, Y. Gao, S. Liu, B. Li, Faraday Discussions 190 (2016) 339–349.
[7] H. Vogel, B. Flerus, F. Stoffner, B. Friedrich, Journal of Sustainable Metallurgy 3, 1
(2017) 99–107.
[8] G.A. Bukatova, S.A. Kuznetsov, M. Gaune-Escard, Russian Journal of
Electrochemistry 43 (8) (2007) 929–935.
[9] P. Sarfo, (2019), Recovery of Rare Earth Elements by Advanced Processing
Technologies, PhD Dissertation, Montana Tech, Butte MT, 120 pp.
[10] E. Stefanidaki, C. Hasiotis, C. Kontoyannis, Electrochimica Acta 46 (17) (2001)
2665–2670.
[11] C. Hamel, P. Chamelot, P. Taxil, Electrochimica Acta 49 (2004) 4467–4476.
[12] R. Seeber, C. Zanardi, G. Inzelt, Chemtexts 2 (2016) 2.
[13] R.W. Balluffi, S.M. Allen, W.C. Carter, Kinetics of materials, Wiley-Interscience,
Hoboken, N.J, 2005.
[14] Y. Shi, X. Liu, Y. Gao, C. Huang, B. Bing Li, Journal of The Electrochemical Society
164 (4) (2017) D263–D268.
[15] J.E. Murphy, D.K. Dysinger, & M.F. Chambers, (1995). Electrowinning Neodymium
Metal from Chloride and Oxide-fluoride Electrolytes. Light Metals, 1313.
[16] G. Knudson & State Univ. of Iowa, Iowa City. (1954). Preparation and properties of
the rare earth fluorides and oxyfluorides. Journal of the American Chemical
Society, 3921-2.
[17] J.D. Hamlin and N. E. Richards. Studies of the Anode Gas from the Hall-Heroult
Cell. In: Paper in Proceedings of 1st International Symposium on the Extractive
Metallurgy of Aluminum. Wiley, 1963, pp.51-63.
[18] P. Sarfo, G. Wyss, G. Ma, A. Das, C. Young, Minerals Engineering 107 (2017) 8–19.
[19] P. Sarfo, A. Das, C. Young, G. Wyss, Waste Management 70 (2017) 272–281.
[20] G.E.P. Box, D.W. Behnken, Technometrics 2 (4) (1960) 455–475.
[21] M.A. Bezerra, R.E. Santelli, E.P. Oliveira, L.S. Villar, L.A. Escaleira, Talanta 76 (5)
(2008) 965–977.
[22] Stat-Ease, Inc. (2003). Design-ease: Software for design of experiments.
Minneapolis, MN: Stat-Ease.
[23] X. Guo, J. Sietsma & Y. Yang, (2016). Chapter 15 - A Critical Evaluation of
Solubility of Rare Earth Oxides in Molten Fluorides.
[24] R. Thudum, A. Srivastava, S. Nandi, A. Nagaraj, R. Shekhar, Mineral Processing
and Extractive Metallurgy 119 (2) (2010) 88–92.
[25] A. Abbasalizadeh, A. Malfliet, S. Seetharaman, J. Sietsma & Y. Yang, (2017).
Electrochemical Extraction of Rare Earth Metals in Molten Fluorides: Conversion of
Rare Earth Oxides into Rare Earth Fluorides Using Fluoride Additives.
[26] X. Hu, Z. Wang, B. Gao, Z. Shi, F. Liu, X. Cao, Journal of Rare Earths 28 (4) (2010)
587–590.