Production of oxygen gas and liquid metal by electrochemical decomposition of molten iron oxide

Article abstract of DOI:10.1149/1.3560477

Molten oxide electrolysis (MOE) is the electrolytic decomposition of a metal oxide, most preferably into liquid metal and oxygen gas. The successful deployment of MOE hinges upon the existence of an inert anode capable of sustained oxygen evolution. Herein we report the results of a program of materials design, selection, and testing of candidate anode materials and demonstrate the utility of iridium in this application. An electrolysis cell fitted with an iridium anode operating at 0.55 A cm^-2 produced liquid metal and oxygen gas by the decomposition of iron oxide dissolved in a solvent electrolyte of molten MgO-CaO-SiO₂-Al₂O₃. The erosion rate of iridium was measured to be less than 8 mm y^-1. The stability of iridium is attributed to a mix of mechanisms including the electrochemical formation and simultaneous thermal decomposition of a surface film of iridium oxide.

Full text of DOI:10.1149/1.3560477

Journal of The Electrochemical Society, 158 (6) E51-E54 (2011)
0
E51
013-4651/2011/158(6)/E51/4/$28.00 VC The Electrochemical Society
Production of Oxygen Gas and Liquid Metal by Electrochemical
Decomposition of Molten Iron Oxide
a,b
Dihua Wang, Andrew J. Gmitter,
a,c,*
a,**,z
and Donald R. Sadoway
a
Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139-4307, USA
School of Resource and Environmental Sciences, Wuhan University, China
b
Molten oxide electrolysis (MOE) is the electrolytic decomposition of a metal oxide, most preferably into liquid metal and oxygen
gas. The successful deployment of MOE hinges upon the existence of an inert anode capable of sustained oxygen evolution.
Herein we report the results of a program of materials design, selection, and testing of candidate anode materials and demonstrate
ꢀ2
the utility of iridium in this application. An electrolysis cell fitted with an iridium anode operating at 0.55 A cm produced liquid
metal and oxygen gas by the decomposition of iron oxide dissolved in a solvent electrolyte of molten MgO–CaO–SiO –Al
2
2 3
O .
ꢀ
1
The erosion rate of iridium was measured to be less than 8 mm y . The stability of iridium is attributed to a mix of mechanisms
including the electrochemical formation and simultaneous thermal decomposition of a surface film of iridium oxide.
VC 2011 The Electrochemical Society. [DOI: 10.1149/1.3560477] All rights reserved.
Manuscript submitted October 25, 2010; revised manuscript received January 31, 2011. Published April 4, 2011.
Modern metals extraction technologies face difficult environmen-
Experimental
9
tal challenges. With worldwide steel production exceeding 1 ꢁ 10
Chemicals and equipment.— Powders of lime, magnesia, alu-
ꢃ
1
tonnes annually, the amount of byproduct carbon dioxide that
mina, and silica (all from Alfa Aesar) were stored at 100 C with
2
accompanies steelmaking (>1.5 tonnes CO per tonne of steel) is
desiccant to prevent adsorption of water and agglomeration. 200 g
batches were dry milled with no milling media in 500 mL Nalgene
jars for roughly 24 h to thoroughly mix. The dry, mixed powders
were poured into an alumina crucible (80 mm OD) or a molybde-
num crucible (60 mm OD). The crucible was set into another alu-
mina crucible and then placed into the bottom of a vertical closed-
one-end alumina tube reactor (113 mm OD ꢁ 105 mm ID ꢁ 750 mm
L, McDanel) in a vertical tube furnace heated by molybdenum disi-
licide elements (Mellen). The electrolyte was blanketed by flowing
argon introduced via a mullite tube positioned in the hot zone of the
noteworthy on the scale of anthropogenic greenhouse gas (GHG)
emissions. The electrolytic production of aluminum also leads to
generation of GHGs in the form of CO , CF , and C F owing to the
2 4 2 6
2
use of a consumable carbon anode (ꢂ1.5 tonnes CO per tonne of
2
aluminum). In response, Sadoway has advocated molten oxide elec-
trolysis (MOE) as a carbon-free alternative to existing metals extrac-
tion technology. MOE is the electrolytic decomposition of a metal
3
oxide, most preferably into liquid metal and oxygen gas. Early efforts
to produce metal by oxide electrolysis go back more than 100 years,
4
but with few exceptions the process has not succeeded in moving
beyond the confines of the laboratory. More recently, variants of
ꢃ
ꢃ
furnace. The furnace was ramped at 85 C=h to 1575 C, and the con-
tents were allowed to equilibrate for roughly one hour prior to elec-
trochemical measurements.
5
MOE have been studied with the intention of exploiting indigenous
resources on the moon and on Mars for the generation of oxygen
along with the production of structural metals such as iron and photo-
Cyclic voltammetry.— The reference electrode was a 3.17 mm
diameter Mo rod shrouded in a mullite tube. The counter electrode
was identically prepared. Two kinds of working electrodes were
used: (1) 0.5 mm diameter Ir wire connected to a 0.368 mm diame-
ter Pt wire; (2) 3.17 mm diameter Mo rod. The exact electrode area
exposed to the electrolyte was determined by a relation between cur-
rent increment and electrode area increment (can be controlled from
the top of the working electrode) at a specific electrode potential.
Cyclic voltammetry with 70% iR compensation was used for all
electrochemical measurements. Higher compensation values
resulted in instability of the system.
6–13
voltaic materials such as silicon.
At this point, the success of
MOE hinges upon the existence of an inert anode capable of sus-
tained oxygen evolution.
Previous work in this laboratory on inert anodes for aluminum-
producing Hall–H e´ roult cells led to the enunciation of a set of selec-
tion criteria for the identification of materials for use as anodes in
1
4
the fluoride-based electrolyte. By analogy, in a molten oxide elec-
trolyte an oxygen-evolving inert anode would desirably be com-
posed of a material with the following attributes:
-
-
-
-
-
-
-
physically stable (preferably solid) at service temperature.
resistant to chemical attack by the molten oxide electrolyte.
resistant to chemical attack by pure oxygen.
electrochemically stable.
electronically conductive.
resistant to thermal shock.
mechanically robust.
Electrolysis and materials characterization.— Constant current
electrolysis was conducted with an electrolyte of composition MgO
(
CaO (18 wt %)–Al O (20 wt %)–SiO (46 wt %) (S1A) to which
26 wt %)–CaO (25 wt %)–SiO
2
(49 wt %) (S1) or MgO (16 wt %)–
2
3
2
was added either 5 or 10 wt % FeO. A Mo rod (3.17 or 6.34 mm in
diameter) was used as the cathode; an Ir plate (5 ꢁ 1 ꢁ 40 mm) or Ir
wire (0.5 mm in diameter, 9.5 mm in length) served as the anode.
The Ir plate was welded to a Mo rod which served as the current
lead. The Ir wire anode was connected to a Pt wire, which served as
the current lead. The Ir that served as the active anode was weighed
before and after electrolysis. The entrained electrolyte on the anode
was removed by immersing it in 47–51% HF aqueous solution. Af-
ter electrolysis the Ir anode and the cathode metal product were
examined by SEM and EDS (JEOL 5910).
In addition, the material should be easy to deploy at industrial
scale; for example, practicalities such as electrical connection to the
bus, cell startup, response to loss of electrical power, etc., should be
straightforward. Application of these selection criteria resulted in
the choice of iridium. The performance of this material as an inert
anode for MOE was assessed in the context of iron production by
electrolysis of FeO dissolved in an alumino-silicate supporting elec-
2 3
trolyte of composition MgO(16 wt %)–CaO(18 wt %)–Al O (20 wt
%
)–SiO (46 wt %).
2
Results
*
*
Electrochemical Society Student Member.
Electrochemical Society Active Member.
Present address: Rutgers University, New Brunswick, NJ.
E-mail: dsadoway@mit.edu
Figure 1 shows the results of cyclic voltammetry on an iridium
wire, 0.5 mm diameter, at oxidizing potentials capable of generating
oxygen gas. The blue trace represents data taken in the supporting
*
c
z
Downloaded on 2014-10-29 to IP 128.248.155.225 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

E52
Journal of The Electrochemical Society, 158 (6) E51-E54 (2011)
Figure 2. (Color online) CVs of a Mo working electrode in molten S1 [SiO
2
(
(
49 wt %)–CaO (25 wt %)–MgO (26 wt %)] with (red line) and without
ꢃ
Figure 1. (Color online) CVs of an Ir wire (diameter 0.5 mm) working elec-
green line) 5 wt % FeO 1575 C in a Mo crucible with a potential scan rate
trode in molten S1A [MgO(16 wt %)–CaO(18 wt %)–Al
2
O
3
(20 wt %)–SiO
ꢃ
2
of 50 mV=s, reference electrode and counter electrode: Mo rods (3.17 mm in
diameter).
(46 wt %)] without (blue line) and with 5 wt % FeO (red line) at 1575 C
with a potential scan rate of 100 mV=s, reference electrode and counter elec-
trode: Mo rods (3.17 mm in diameter).
In contrast, the red trace representing the voltammogram of the elec-
trolyte containing 5 wt % FeO shows a strong reduction current on
the cathodic sweep at potentials as high as –0.05 V, in excellent
agreement with the –0.08 V value from thermodynamic calculation,
electrolyte (S1A), while the red trace represents data taken in the
same melt to which 5 wt % FeO has been added. At potentials less
2þ
d
=Mo) , the blue trace shows near zero current
and in conformity with the electroreduction of Fe
Galvanostatic electrolysis was performed at a current density of
.
than 0.75 V (vs. MoO
3
indicating that iridium is electrochemically stable and that the elec-
tronic conductivity of the melt is negligible. At potentials exceeding
ꢀ
2
.55 A cm for a period of 6.35 h on a melt composed of the
0
alumino-silicate supporting electrolyte containing 5 wt % FeO. An
iridium wire, 0.5 mm diameter, 9.5 cm long, immersed 1.4 cm into
the melt, served as the anode, while a molybdenum rod, 3.17 mm
diameter, served as the cathode. The cell was also fitted with a refer-
ence electrode consisting of a molybdenum rod, 3.17 mm diameter.
Temperature was set at 1575 C. The black trace of Fig. 3 shows that
the potential on the anode as measured between it and the reference
electrode remained nearly constant. The condition of the iridium
1
.05 V the current grows measurably, and exhibits significant fluctu-
ation on the order of 10 s of millivolts, which is consistent with gas
evolution, i.e., formation, coalescence, and detachment of bubbles.
In addition, there is a feature between 0.75 and 1.05 V, for which
the peak current was found to be scan-rate dependent. This feature
is assigned to the electrochemical formation of iridium oxide, which
will be discussed below. Such putative behavior is similar to that
observed during anodic oxidation of platinum in molten silicates as
ꢃ
15,16
reported by Higgins
where he suggests the formation of an in-
termediate platinum oxide. The red trace in Fig. 1 shows that even
upon addition of FeO to the alumino-silicate supporting electrolyte
the electrical signatures of both oxygen evolution and putative irid-
ium oxide formation are present. The positive slope in the current
response at potentials lower than 0.75 V may be due to increasing
2
þ
electronic conductivity of the melt or oxidation of some Fe which
2þ
3þ
could result in polaron hopping between Fe and Fe . The higher
oxygen evolution current is attributed to the greater concentration of
free oxide ions in the FeO-bearing melt.
Figure 2 shows the results of a voltammetric study of iron deposi-
ꢃ
tion at 1575 C using a molybdenum rod working electrode. The
green trace represents data taken in the supporting electrolyte (S1),
while the red trace represents data taken in the same melt to which 5
wt % FeO has been added. The electrochemical window of the sup-
porting electrolyte extends in the cathodic direction to –0.6 V at
which potential the observed current is assigned to the electrodeposi-
1
7
tion of liquid silicon. Since, on thermodynamic grounds, the onset
of silicon deposition should occur at potentials roughly 0.3 V more
cathodic than observed in Fig. 2, the data indicate underpotential
deposition, which is not surprising given that silicon and molybde-
1
8
num readily alloy at this temperature, as do iron and molybdenum.
In the anodic direction, an oxidation current is observed at potentials
exceeding 0 V and is assigned to the oxidation of molybdenum itself.
Figure 3. (Color online) Variation of potential with time of an Ir wire anode
(
black line, 0.5 mm diameter, 1.4 cm in melt) in S1A with 5 wt % FeO at
ꢃ
2
575 C and of an Ir plate anode (red line, area: 2.52 cm ) in S1A with 10 wt
1
% FeO at 1575 C during constant current electrolysis.
d
ꢃ
All potentials are reported against a MoO
3
=Mo pseudo-reference electrode.
Downloaded on 2014-10-29 to IP 128.248.155.225 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 158 (6) E51-E54 (2011)
E53
Figure 4. SEM image of original Ir wire anode (0.5 mm
in diameter) before (a) and after (b) electrolysis for 6.35 h
ꢃ
in S1A þ 5 wt % FeO at 1575 C with an anodic polariza-
ꢀ
2
tion current density of 0.55 A cm
.
anode is depicted before and after electrolysis (the entrained electro-
lyte was removed by immersing the iridium wire in 47–51% HF
aqueous solution for 45 min) in Figs. 4a and 4b, respectively, show-
ing a change in surface finish but no perceptible change in diameter.
The mass of the iridium wire was 440.13 and 437.12 mg before and
after electrolysis, respectively. The 3.01 mg weight loss of the entire
iridium wire is equivalent to a consumption rate of ꢂ7.7 mm=year
when ascribed only to the section of wire immersed in the electro-
lyte. According to the standards set forth by the aluminum industry
At the anode, we speculate that the mechanisms at work include the
following, where the term, melt, refers to the molten oxide electrolyte
- electrochemical oxidation of iridium to form iridium oxide (iri-
dia), which is a gas at this temperature
2
ꢀ
ꢀ
Ir þ y O ! Ir OyðgÞ þ 2y e ; ðy ¼ 1:5; 2; or 3Þ
Ir þ y O2ꢀ ! Ir OyðsÞ þ 2y eꢀ; ðy ¼ 1:5; 2; or 3Þ
[3]
0
[3 ]
19
(
less than 10 mm=year, the only industrial specification available )
- dissolution of iridia in the molten oxide electrolyte
IrOyðgÞ þ MOzðmeltÞ ! MIrOyþzðmeltÞ
IrO ðgÞ ! IrO ðmeltÞ
such a wear rate qualifies for the designation of inert anode. How-
ever, if one accounts for anode consumption above the melt line by
thermochemical oxidation of iridium and sublimation of the iridium
[4]
0
20
[4 ]
y
y
oxide so formed, the 7.7 mm=year figure can be viewed as an
upper bound, i.e., the erosion rate of the anode immersed in the melt
is less than this value. Subsequent testing of an anode made of irid-
ium plate, 40 ꢁ 5 ꢁ 1 mm, confirmed the initial findings. In this sec-
ond experiment the electrolyte contained 10 wt % FeO dissolved in
the alumino-silicate supporting electrolyte. Cell voltage was seen to
be stable even as current density was varied from 0.60 to 1.19 A
-
decomposition of iridia into its constituent elements
IrOyðgÞ ! IrðelectrodeÞ þ 0:5y O2ðgÞ
IrOyðgÞ ! Ir ðdispersed in the meltÞ þ 0:5yO2ðgÞ
- electrochemical dissolution of iridium, i.e., electrostripping
[5]
0
[5 ]
ꢀ
2
cm as seen in the red traces of Fig. 3. After electrolysis the plate
thickness was visibly unchanged, and no iridium was detected by
EDS and XPS in the cathode product or in the electrolyte frozen on
anode. The bulk electrolyte far from the anode was not searched for
evidence of elemental iridium. The metal on the cathode was deter-
mined by EDS to be an alloy of iron and molybdenum. Electro-
chemical generation of oxygen was confirmed by chemical analysis
of gas evolving on the anode (Model O2 Oxygen Analyzer, Oxigraf,
Inc., Mountain View, CA) and swept away in an argon flow. By the
same method no oxygen was detected when no current passed
through the cell.
Ir ! Ir þðmeltÞ þ 4eꢀ
4
[6]
In Fig. 1, we see in the feature between 0.75 and 1.05 V possible
evidence for Reaction 3 or Reaction 6. If Reaction 3 is operative
then the fate of such electrochemically generated iridia on the anode
surface determines the utility of the anode. Any of the following
reactions would consume iridia and thereby deplete the anode of
0
material: dissolution (Reactions 4 and 4 ), decomposition (Reaction
0
5 ), or fugitive emission (Reaction 3). The exception is Reaction 5
which does not result in the removal of iridium from the anode. In
fact, if the rate of thermal decomposition of iridia according to
Reaction 5 occurs at a rate that exceeds the rate of iridia generation
by Reaction 3, this will result in the continuous release of oxygen
gas which bubbles away and the continuous release of iridium metal
which remains on the anode. Evidence for the latter can be seen in
Fig. 4b which reveals surface reconstruction of the iridium without
loss of material. Such an iridium electrode would be termed an iri-
dia-mediated inert anode (IMIA), or more generally an oxide-medi-
ated inert anode (OMIA). Alternatively, if k ꢄk , where i ¼ 3–6, we
Discussion
On the basis of these experimental data we put forth the follow-
ing scheme to explain the performance of iridium as an oxygen-
evolving inert anode in iron-bearing alumino-silcate melts. The
overall half-cell reactions during the electrolysis of FeO are
cathode :
anode :
Fe þ 2eꢀ ! FeðlÞ
2
þ
[1]
[2]
2
i
2ꢀ
! =2O2ðgÞ þ 2eꢀ
1
O
should expect the same outcome.
Figure 5. (Color online) (a) Photo image showing de-
posit on a hanging Mo cathode (6.3 mm in diameter) af-
ter electrolysis of S1A melt containing FeO and (b) SEM
image of deposit and EDS analysis (superimposed in
red) of same.
Downloaded on 2014-10-29 to IP 128.248.155.225 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

E54
Journal of The Electrochemical Society, 158 (6) E51-E54 (2011)
The concept of an oxide-mediated inert anode should not be
sions. The authors thank Guenther Arndt for machining the anodes
and dissecting the crucibles. The authors thank Drs. Antoine Alla-
nore, Hojong Kim, and Luis A. Ortiz for their careful review of the
manuscript and many suggestions for improvement of same.
limited to iridium metal and molten oxide melts; in principle, the
anodic behavior of other noble metals including Pt, Rh, Au, and Ag
may be explained to the extent that they form an intermediate com-
pound by chemical combination with a constituent of the electro-
lyte in conformity with the set of reactions set forth above. Thus,
one might possibly be able to obtain an inert anode by adjusting
relative rates of Reactions 3–6 through control of temperature, elec-
trolyte composition, and metal alloy composition. Given that in ha-
lide melts oxygen evolution occurs at less extreme potentials than
does halogen evolution, we speculate on the potential utility of an
oxide-mediated inert anode for other metallurgical processes of
interest.
Massachusetts Institute of Technology assisted in meeting the publication
costs of this article.
References
1. World Steel in Figures, World Steel Association, Brussels (2010).
2
3
4
5
.
.
.
.
E. Worrell, L. Price, and N. Martin, Energy (Oxford, U. K.), 26, 513 (2001).
D. R. Sadoway, J. Mater. Res., 10, 487 (1995).
R. H. Aiken, U.S. Pat. 816142 (1906).
R. Winand, A. Fontana, L. Segers, P. Hannaert, and J. Lacave, in Molten Salt Elec-
trolysis Metal Production International Symposium, The Institution of Mining and
Metallurgy, London, pp. 42–50, (1977).
In summary, by presenting results for electrochemical decompo-
sition of FeO dissolved in an alumino-silicate melt to produce
ꢃ
oxygen gas and liquid metal at 1575 C we have demonstrated that
6.
R. O. Colson and L. A. Haskin, in Resources of Near-Earth Space, J. S. Lewis, M.
S. Matthews and M. L. Guerrieri, Editors, pp. 109–27, University of Arizona Press,
Tucson (1993).
iridium can serve as an oxygen-evolving anode in an oxide melt at
extremely high temperature. Several mechanisms have been put
forth to explain the performance of iridium and to provide guidance
to future efforts at inert anode development in other electrolytic
systems. While iridium is far too costly and its natural abundance
in the earth’s crust is far too low to meet the needs of the world
steel industry, iridium has an important role to play in laboratory-
scale studies of electrolytic cells evolving oxygen and in special-
ized settings such as in situ generation of oxygen in extraterrestrial
venues.
7. D. G. Kesterke and S. P. Nasa, NASA Spec. Publ. (NASA SP-229), pp. 139–44
1970).
(
D. G. Kesterke, U.S. Bur. Mines, Rep. Invest. (RI 7587), p. 12 (1971).
D. Khetpal, A. C. Ducret, and D. R. Sadoway, NASA Conf. Publ., 2003–212339,
8
9
.
.
548 (2003).
10. D. J. Lindstrom and L. A. Haskin, Lunar Planet. Sci., 10, 733 (1979).
11. M. J. Oppenheim and H. Tabor, Adv. Space Sci. Technol., 11, 233 (1972).
12. A. H. C. Sirk, D. R. Sadoway, and L. Sibille, ECS Trans., 28(6), 367 (2010).
13. A. T. Vai, J. A. Yurko, D. H. Wang, and D. R. Sadoway, in Jim Evans Honorary
Symposium, B. Q. Li, B. G. Thomas, L. Zhang, F. M. Doyle, and A. P. Campbell,
Editors, pp. 301–308, TMS, Warrendale, PA (2010).
Acknowledgments
14. D. R. Sadoway, Light Metals 1990: Advances in Aluminum Production, C. A.
Bickert, Editor, pp. 403-07, TMS, Warrendale, PA (1990).
The authors are grateful to the American Iron and Steel Institute
AISI) for financial support. DHW thanks the Chinese Scholarship
Council (CSC) for affording him the opportunity to work at MIT
and Dr. Bing Li for her assistance and discussion during bench
work. AJG thanks Dr. Aislinn Sirk who entertained helpful discus-
15. J. K. Higgins, Glass Technol., 21, 145 (1980).
1
1
1
1
6. J. K. Higgins, Glass Technol., 23, 90 (1982).
(
7. A. Roine, HSC Chemistry 6.0, Outokumpu Technology, Pori, Finland (2006).
8. A. B. Gokhale and G. J. Abbaschian, J. Phase Equilib., 12, 493 (1991).
9. Inert Anode Roadmap, U.S. Department of Energy, Washington, DC (1998).
20. H. Jehn, R. Voelker, and M. I. Ismail, Platinum Met. Rev., 22, 92 (1978).
Downloaded on 2014-10-29 to IP 128.248.155.225 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).