Design of a Molten Materials Handling Device for Support of
Molten Regolith Electrolysis
THESIS
Presented in Partial Fulfillment of the Requirements for the Degree of Master of
Science in the Graduate School of The Ohio State University
By
Evan Standish
Graduate Program in the Department of Materials Science and Engineering
B.S. Materials Science and Engineering
The Ohio State University
2010
Masters Examination Committee:
Dr. Doru Stefanescu “Advisor”, Dr. Yogeshwar Sahai
Copyright by
Evan C Standish
2010
Abstract
This study was performed to develop a method of removing molten process fluids, namely a
ferrosilicon alloy and a complex silicate melt, from an electrolysis cell. The device was designed
as a component of equipment used for an in situ lunar oxygen generating process under
development. This work focuses on developing a system for integration into a molten regolith
electrolysis cell, the products of which are highly reactive, making materials compatibility a
primary concern of a materials handling system. This paper describes the design and operation of
a mechanism utilizing a pressure differential to pull molten material from a furnace into a mold.
The results of several different materials choices for equipment hardware are described and
suggestions for modification of the device for improvement and lunar compatibility are made.
ii
Acknowledgments
I would like to acknowledge and thank:
NASA for financial sponsorship of the research.
Professor Doru Stefanescu for his guidance and always open door.
Professor Jerrald Brevick for providing background on casting processes and helping to
brainstorm ideas.
Professor Yogeshwar Sahai for his review of this document.
Dr. Laurent Sibille and the NASA management team for their support of the project.
Mr. Ken Kushner, Mr. Ross Baldwin, and Mr. Gary Dodge for providing assistance with use of
departmental facilities.
Mr. Martin Moneysmith for advice on the design of the vacuum dome.
Mr. Andrew Gledhill for his help with performing ceramic corrosion tests.
Mr. Sarum Boonmee and Mr. Bobby Gyesi from the Stefanescu lab for providing assistance with
performing experiments.
Ms. Katie Shortman for always being interested in hearing about my work and helping with editing
of this paper.
My parents for their continual support of all my activities, especially their promotion of my
scientific interests whether at home in the kitchen at school or in the lab.
iii
Vita
June 2005 .......................................................... Hudson High School
June 2009 .......................................................... B.S. Materials Science and Engineering, The
Ohio State University
March 2008 to present ...................................... Graduate Research Associate, Materials
Science and Engineering Department, The Ohio
State University
Fields of Study
Major Field: Materials Science and Engineering
iv
Table of Contents
Abstract .......................................................................................................................... ii
Acknowledgments ......................................................................................................... iii
Vita ............................................................................................................................... iv
Table of Contents ............................................................................................................v
List of Figures ............................................................................................................... vi
List of Tables............................................................................................................... viii
1 Introduction ..............................................................................................................1
1.1 Background .......................................................................................................1
1.2 Objectives of Project .........................................................................................3
2 Experimental Strategy ..............................................................................................1
3
Ceramics Investigation
.............................................................................1
3.1 Explanation of Containment Difficulty ..............................................................1
3.2 Experimental Procedure.....................................................................................3
3.3 Results...............................................................................................................4
3.4 Discussion .........................................................................................................8
3.5 Conclusions ..................................................................................................... 11
4
Design of the Experimental Apparatus .....................................................................1
4.1 Design Constrains of the Fluid Removal Device ................................................1
4.2 Current Methods for Withdrawal of Molten Fluids and Method Selection ..........1
4.3 Description of the Experimental Apparatus ........................................................4
4.4 Evolution of Design ......................................................................................... 12
4.5 Limitations of Current Design and Scale-up Issues .......................................... 13
4.6 Testing Strategy............................................................................................... 14
5 Experimental Withdrawal Procedures and Results ....................................................1
5.1 Water ................................................................................................................1
5.2 Corn Syrup ........................................................................................................5
5.3 Aluminum .........................................................................................................5
5.4 Ferrosilicon ..................................................................................................... 12
5.5 Low Viscosity Glass ........................................................................................ 17
5.6 Depleted Regolith ............................................................................................ 22
5.7 Dual Withdrawal of Molten Regolith and Ferrosilicon ..................................... 27
6 General Discussion and Thoughts on Lunar Implementation.....................................1
6.1 Focus Areas for Further Development ...............................................................1
6.2 Process Modeling ..............................................................................................3
6.3 General Re-Design Considerations ....................................................................4
7 Conclusions ..............................................................................................................1
Appendix A: Experimental Geometry ..............................................................................1
Appendix B: Additional Experimental Data .....................................................................2
References .......................................................................................................................3
v
List of Figures
Figure 1: Plot of reduction potentials and melting point of oxide electrolyte during MRE process ........... 2
Figure 2: Corrosion specimen after removal from furnace, numbers correspond to Table 4. These
samples were held for six hours in air. (Scale reference: side lengths of samples 11, 12 are 2.54 x
10-2 m (one inch)). ........................................................................................................................... 5
Figure 3: Specimen after heating to temperature in argon and immediate cooling; no hold was
performed. (Scale reference: side lengths of samples 11, 12 are 2.54 x 10-2 m (one inch)). ............. 5
Figure 4: Alumina sample #1 after an eight hour hold in air. Inset shows the untested structure of the
ceramic for comparison. No penetration by the regolith simulant is apparent................................ 6
Figure 5: Boron nitride sample # 11 from Figure 3. Left shows interface reaction product, right shows
BSE SEM picture of same region; labeled areas one and two are identified by EDS in Table 6. ........ 6
Figure 6: Cross sections of experimental specimen. Numbers correspond to conditions in ...................... 7
Figure 7: Cold walled cell schematic. ........................................................................................................ 1
Figure 8: Mechanisms for closing gravity operated fluid removal methods. Left: slide gate, Center:
tapered plunger, Right: Ceramic plug in burned hole. ..................................................................... 2
Figure 9: Counter gravity materials removal. Pressure differential can be created either by applying
pressure on the melt or by pulling a vacuum over the removal tube. ............................................. 3
Figure 10: General dimensions (mm) of the sand mold used to receive the withdrawn melt. The inside
volume up to the ledge is 283 ml. ................................................................................................... 6
Figure 11: Temperature-viscosity curves for compositions similar to a depleted regolith melt. ............... 8
Figure 12: Temperature-viscosity curve estimate for 60wt% Si 40% Fe. Viscosity obtained by mol
fraction average of Si and Fe metals. Viscosities below melting points obtained by linear
extrapolation ,. ................................................................................................................................ 9
Figure 13: Boron nitride coated molybdenum tube. Notch for directing outlet flow on right. Uncoated
portion allows better gluing. ......................................................................................................... 10
Figure 14: Pressurized withdrawal requires a dynamic high temperature seal against the tube if
translation is to be achieved during withdrawal. Using vacuum withdrawal eliminates the need for
a dynamic seal because the pressure inside and outside the furnace (electrolysis cell) is the same.
...................................................................................................................................................... 11
Figure 15: Initial (left) and final (right) configuration used for the differential pressure withdrawal
device. .......................................................................................................................................... 13
Figure 16: (Continued) Pressure is reduced inside the dome allowing atmospheric pressure to push fluid
up the withdrawal tube (1). As the mold fills the level of liquid in the crucible drops until it
reaches the inlet of the tube, or the critical height, at this point bubbles are ingested in the tube.
(2,3). Once vacuum is lost the fluid in the tube falls back into the crucible raising the height of
liquid, once again sealing the tube (4). Upon resealing the pressure differential is rapidly re-
vi
achieved causing the uncontrolled rise of liquid up the withdrawal tube and splashing into the
vacuum chamber (5). ...................................................................................................................... 2
Figure 17: Insulation covers resistance heating wire in Trial 5. ................................................................. 6
Figure 18: Withdrawal tube was glued into sodium silicate bonded sand with alumina paste. Alumina
fiber was used to insulate the tube. ................................................................................................ 7
Figure 19: Frozen aluminum in silica withdrawal tube showing only one inch of removal past the initial
melt height...................................................................................................................................... 9
Figure 20: Glass dome after cracking due to thermal shock from molten aluminum splash. .................. 10
Figure 21: Ferrosilicon immediately after withdrawal in the second experiment. .................................. 15
Figure 22: Results of low viscosity glass withdrawal. Mass around exit of tube had a high viscosity
during withdrawal while the glass in the mold flowed easily. This difference is probably due to
chilling of the initial material by the tube. .................................................................................... 20
Figure 23: Pressure and temperature data from low viscosity glass withdrawal. Pressure measurement
is absolute pressure in vacuum chamber; temperature was recorded at two inches below the top
of the tube in between the sand mold and the tube. Synchronization of the data is approximate
because two different boards were used for signal acquisition. Time boundary is arbitrary; zero is
the start of data acquisition. ......................................................................................................... 21
Figure 24: Temperature data recorded from tube thermocouples, melt thermocouple did not yield any
significant reading. The fluctuations in thermocouple output began around time = 175 seconds,
right as the lower tube thermocouple was broken, it is likely that this caused interference with
the data acquisition board. Time is zeroed to first downward motion of the actuator. Withdrawal
occurs at time = 170 seconds......................................................................................................... 24
Figure 25: Pressure profile from depleted regolith experiment compared to that of the low viscosity
glass. The low viscosity glass data has been translated to overlay. ............................................... 25
Figure 26: Depleted regolith withdrawal results. ................................................................................... 26
Figure 27: Cross section of withdrawn ingot from Trial 1 showing both ferrosilicon layer and an oxide
layer. ............................................................................................................................................. 30
Figure 28: Schematic diagrams of potential high temperature dynamic sealing options. An additional
concept not presented is the use of an edge welded bellows to achieve a small degree of
translational freedom. .................................................................................................................... 2
Figure 29: Cross section of furnace and vacuum chamber. Several dimensions, (mm), are presented for
scale. ............................................................................................................................................... 1
Figure 30: Pressure from dual withdrawal Trial #1. .................................................................................. 2
Figure 31: Temperature measurements from dual withdrawal Trial #4. ................................................... 1
Figure 32: Temperature profile from dual withdrawal Trial #5. ................................................................ 1
Figure 33: Absolute pressure in vacuum chamber during dual withdrawal Trial #5. ................................. 2
vii
List of Tables
Table 1: Chemical composition (major components) of lunar regolith ..................................................... 1
Table 2: Composition of depleted regolith oxide mixture......................................................................... 2
Table 3: Binary eutectics of various components of the potential regolith-crucible system (°C) ............... 1
Table 4: Ceramics evaluated in preliminary compatibility trials. .............................................................. 3
Table 5: Corrosion experimental melt compositions. ............................................................................... 4
Table 6: Standardless EDS results for corrosion product shown in Figure 5. ............................................. 7
Table 7: Conditions of experiment from Figure 6. .................................................................................... 7
Table 8: Pressure differentials for constant temperature, steady state withdrawal in earth gravity. ....... 9
Table 9: Significant modifications between initial and final withdrawal system configuration ............... 12
Table 10: Aluminum Experimental Setup ................................................................................................. 6
Table 11: Aluminum Experimental Parameters ........................................................................................ 8
Table 12: Aluminum Experimental Results Summary ............................................................................... 8
Table 13: Ferrosilicon Experimental Setup ............................................................................................. 12
Table 14: Ferrosilicon Experimental Parameters .................................................................................... 13
Table 15: Ferrosilicon Experimental Results Summary ........................................................................... 13
Table 16: Depleted Regolith vs. Modified Low Viscosity Melt Composition............................................ 17
Table 17: Low Viscosity Glass Experimental Setup ................................................................................. 17
Table 18: Low Viscosity Glass Experimental Parameters ........................................................................ 19
Table 19: Low Viscosity Glass Experimental Results Summary ............................................................... 19
Table 20: Depleted Regolith Experimental Setup ................................................................................... 23
Table 21: Depleted Regolith Experimental Parameters .......................................................................... 23
Table 22: Depleted Regolith Experimental Results Summary ................................................................. 24
Table 23: Dual Withdrawal Experimental Setup ..................................................................................... 28
Table 24: Dual Withdrawal Experimental Parameters ............................................................................ 29
Table 25: Dual Withdrawal Experimental Results Summary ................................................................... 29
viii
Section 1.
Introduction
1.1 Background
As part of a National Aeronautics and Space Administration (NASA) objective to return to the
Moon and establish a permanent presence, a means of producing in situ oxygen from lunar
materials is a necessary goal. In previous lunar missions every consumable resource available to
astronauts was brought from Earth at great expense. With NASA‟s current interest in developing
long term exploration leading to permanent lunar habitation there will be a great need for oxygen
availability on the Moon for a variety of purposes. In addition to life support, oxygen could be
useful to astronauts and lunar based devices for power generation and storage via fuel cells,
water generation, combined with Earth sourced hydrogen, and as an oxidizer for rocket based
propulsion. With a lunar source of oxygen much of this consumable mass could be eliminated for
considerable cost savings to a lunar exploration and habitation program.
Given the great benefit of lunar sourced oxygen, it has been a research objective for decades to
develop a means by which oxygen gas could be produced from lunar materials. Many research
projects have been directed at producing oxygen from the lunar soil, known as regolith, which is
composed of various oxides ranging in particle size from large rocks to fine powder. A summary
of the processes proposed is provided by Taylor and Carrier1. Though this paper is somewhat
outdated, two of the top processes identified are still the favorites today. One method, gaseous
reduction, uses a gas; hydrogen is the favorite though methane or carbon monoxide also work, to
reduce the oxygen from a crushed regolith powder in a fluidized bed reactor. The gaseous
product is then electrolyzed using solar power to produce oxygen and regenerate the reducing
agent. A third method being developed is the molten regolith electrolysis (MRE) process in which
regolith is heated to its melting point and current is passed through the melt to reduce the
constituent oxides to metal and create free oxygen gas. A summary of this technology is given by
2
Curreri .
Among these methods, the second is seen as best fitting the objectives of the In Situ Resource
Utilization (ISRU) program at NASA which seeks to maximize use of available resources on the
lunar surface with as few consumable materials from the Earth as possible. The benefits to the
1
MRE process include the potential for elimination of consumable materials, a higher oxygen yield
from lunar materials, and a greater number of usable products. In the gas phase reduction
process though hydrogen is largely recycled some of the gas inevitably escapes to the lunar
vacuum during processing and recovery and would require periodic replacement. Additionally the
process has a low oxygen yield, resulting in low energy efficiency given that the remaining portion
of the heated mass is not recovered as a useful product. Beneficiation of the ore used in the
reactor increases the yield, but requires an additional step. The MRE process however allows a
recovery of up to 35% of the available oxygen in lunar regolith without requiring any beneficiation
and in addition to oxygen yields a molten ferrosilicon alloy, a highly useful product, readily
adaptable to further processing.
The development of molten regolith electrolysis has been under investigation for many years; it
has also been known by the name of molten silicate electrolysis, and magma electrolysis
depending on the particular study. The latest investigations have been ongoing for ten years at
Massachusetts Institute of Technology, (MIT). Work on the process has been part of a wider
investigation into generalized molten oxide electrolysis conducted under the supervision of
Donald Sadoway. The current state of the art involves melting lunar regolith simulant JSC-1A and
heating to a process temperature of 1600 °C where electricity is passed between a molybdenum
cathode and a largely inert iridium anode. At the cathode, the oxide constituents of the melt are
reduced starting with the alkali components and progressing to the more active metals. Figure 1
depicts the melting point of the electrolyte as electrolysis progresses. The melting point of the
oxide electrolyte increases until electrolysis is stopped just before the designated stopping point
of a 1600 °C liquidus. At this point the composition of the electrolyte melt is similar to that shown
in Table 2.
Table 1: Chemical composition (major components) of lunar regolith3
Material
Oxide
SiO2
TiO2
Al2O3
Fe2O3
FeO
MgO
CaO
Na2O
K2O
MnO
Cr2O3
JSC-1
Lunar Soil
(mean of 3)
14163*
Concentration Std. Dev. Concentration
Wt %
Wt %
Wt %
47.71
0.1
47.3
1.59
0.01
1.6
15.02
0.04
17.8
3.44
0.03
0
7.35
0.05
10.5
9.01
0.09
9.6
10.42
0.03
11.4
2.7
0.03
0.7
0.82
0.02
0.6
0.18
0
0.1
0.04
0
0.2
1
P2O5
0.66
LOI
0.71
Total
99.65
LOI = Loss on ignition
0.01
0.05
----99.8
Figure 1: Plot of reduction potentials and melting point of oxide electrolyte during MRE process4
Table 2: Composition of depleted regolith oxide mixture.
Compound
SiO2
Al2O3
MgO
CaO
FeO
Total
Wt Percent
37.9
28.5
15.3
18.3
0.0
100.0
Analysis of the thermodynamics of the melt constituents shows that for the prescribed
composition the two primary cathodic reactions of the process are reduction of iron and silicon4.
FeO(l )
2e
Fe(l ) [O 2 ]
(1.1.1)
SiO2(l )
4e
Si(l )
2[O 2 ]
(1.1.2)
The anodic reaction is the generation of gaseous oxygen which bubbles out of the cell and is
captured.
2
MOx (l )
[M 2 x ]
x
O2( g ) 2 xe
2
Much of the previous work on MRE has been related to determining performance at different
temperatures and choosing electrode materials. Establishing an appropriate temperature is
important to define the physical properties of the electrolyte and for controlling the chemical
reactions which can take place in the aggressive environment of a molten oxide electrolysis cell.
Determining a material which can act as an inert electrode in the severe chemical environment of
the cell has been a challenge which is still under investigation at MIT and Kennedy Space Center,
(KSC).
In addition to electrochemical research, development of the MRE process requires consideration
into the physical operation of the electrolysis cell. This is something which at first glance seems
like an easy problem to solve, given that similar electrowinning operations are in use for
production of various reactive metals. But under the constraints of very high temperature, reactive
materials containment and handling, and the inability to use consumable materials, it becomes
challenging to determine an appropriate cell design. Much of the current understanding of
industrial scale electrochemical reduction processes comes from the aluminum industry where
the electrolytic production of aluminum from aluminum oxide has been practiced for over 100
years5. It is from the Hall-Heroult method and cell design that the conceptual design of a molten
oxide electrolysis cell has been derived.
One challenge which is somewhat unique to the MRE process is the method by which the
products of the cell are removed. In typical earth bound systems where the focus of
electrowinning cells is the metallic product it is not important to capture any gaseous products of
reaction. However, for MRE oxygen is the primary target and so any system which is used to
withdrawal molten materials from the cell must also ensure the ability of containment and capture
of the oxygen gas evolved from the cell. The main focus of this project is the design and
operation of a device which can be used for removal of molten oxides and molten metals from an
electrolysis cell.
1.2 Objectives of Project
The primary objective of this project is the design and implementation of a method and any
included devices for removing molten materials from a molten regolith electrolysis cell. In lunar
oxygen production a range of feedstocks are possible and process variations are optional, this
means that a range of products can be produced. The list of possible materials to be removed
includes iron, an iron-silicon alloy and molten glasses of a range of compositions. However, the
majority of electrolysis experiments have been conducted on JSC-1A lunar simulant and so the
focus of this project has been on the development of a system which is most effective for the
3
(1.1.3)
1
likely products of that chemical system, namely a 70 wt% silicon, 30 wt% iron alloy and a glass
with a composition as shown in Table 1.
A secondary objective of this experiment was to evaluate various ceramic components for their
compatibility with molten oxide for improving performance of containment devices during
electrolysis trials. Information obtained in this investigation was also used to make materials
selection choices for fluid handling components.
1
The process temperature is above the boiling point of the alkali components in the melt so these are
not expected to exist in the metallic product alloy. To simplify experiments trace oxides are ignored.
4
Section 2.
Experimental Strategy
The first objective addressed in this project was that of materials compatibility. At the start of the
project the issue of containing the oxide electrolyte was of critical and immediate concern to
experimenters at MIT. Difficulties in containing the electrolyte during high temperature
experiments led to multiple crucible failures and prevented experiments from yielding conclusive
results. Any improvements in material selection for containment of oxide melts could be
implemented immediately in the electrolysis experiments, reducing concerns about interrupted
experiments and damage to expensive furnace hardware.
Additionally, the design of a withdrawal device required understanding of the limitations of
construction materials. A typical design approach would be to define constraints based on the
mechanics of the device and chose a material to suit the needs of an established design.
However, in the case of a molten regolith electrolysis cell the environment is so chemically
reactive that it was decided to take the opposite approach and design the device around a limited
set of possible materials. This approach was taken with the understanding that it would likely be
easier to design a device around an existing material than try to invent a new material solution to
fit the needs of a design which was materials compatibility limited.
Following the identification of a list of compatible materials solutions, the next step of the project
was to narrow down from a short list, a method of removing molten materials from the reactor. A
summary of possible methods was completed by Stefanescu 6. Discussion of the process for
making this selection is presented in section 4.2 of this thesis. Upon selection of a method,
design of the system and construction of the initial configuration were begun. Section 4.3.2,
describes the design considerations for the device and section 4.4 compares the initial
configuration to the final result. Following completion of the initial design, testing of the device
using progressively more demanding conditions commenced. As experiments progressed
modifications to the design were constantly made to allow improved performance. Finally after
successfully proving the feasibility of the method and equipment several materials variations for
critical components were tested to compare results under similar withdrawal conditions. This work
is described in section Section 5. Section Section 6 of this thesis describes the work that needs to
be done to modify the design of this device to allow integration into a cell for use in a lunar
environment.
1
Section 3.
Ceramics Investigation
3.1 Explanation of Containment Difficulty
While the composition of lunar regolith varies based on location, a good average is given by
simulant JSC-1A which has a chemical composition as described in Table 1. A complete
description of the simulant is given by Papke et al7, but only the chemical nature of the material is
of concern for this work since any material to be handled will be melted first.
Based upon the reduction potentials of the component oxides in the electrolyzed regolith it is
expected that the process should generate three metallic components, silicon, iron and a small
amount of sodium. After reduction of the metallic components, approximately 60 wt% of the initial
charge will remain as a slag to be removed from the cell. This dictates that any material used to
contain the process be able to withstand contact with three distinctly substances: a glassy silicate
melt, a molten iron-silicon alloy, and high temperature oxygen gas at 1600 °C. This is a
challenging proposition as the materials produced in the process are highly interactive with
different containment materials. Given the temperatures involved, the two reasonable choices for
containment are either ceramics or refractory metals such as tungsten, niobium or molybdenum.
The use of the metals as a containment vessel however is excluded by the fact that one of the
products of the electrolysis reaction is a molten iron-silicon alloy. The solubility of the refractory
metals in iron will induce dissolution and failure of the crucibles. Additionally, the refractory metals
are known to oxidize readily at high temperatures, and exposure to the highly oxidizing
environment created in a molten regolith electrolysis cell would rapidly destroy any metallic
components. Accordingly this work was focused on ceramics as long term, (on the order of 10-20
hours at temperature), containment materials for electrolysis experiments. A qualitative approach
to evaluating compatibility was taken, as the identification of chemical reactions and mechanisms
of corrosion was not the objective of this project.
3.1.1 Oxide Ceramics
Ceramics cover a wide range of materials with highly varying properties, being generally
described as hard, brittle and possessing a high melting point. The two major classes of ceramics
are oxides and non-oxides. Those in the oxide category are used in bulk in many industries and
have been used since antiquity for applications requiring temperature resistance. The non-oxides,
which include carbides, nitrides, and borides, are still
1
being actively developed and are used primarily in high-tech applications in which their superior
performance for specific applications outweighs their high cost. Given their availability and
relatively low cost, the first choice for ceramic containment equipment would be the oxide
ceramics. However, because the electrolyte in molten regolith electrolysis is itself a mixture of
molten oxides, it is questionable whether an oxide crucible would withstand the aggressive nature
of the reactant for long periods of time. The major component of lunar regolith, silica, is regarded
as a universal solvent of other oxides so it is likely that any oxide ceramics used will exhibit some
solubility in the melt. Additionally, the many lower mass fraction components of lunar regolith
provide the possibility of incompatibility with a wide range of candidate oxides. It is necessary
therefore to determine a procedure for choosing the most likely candidates and to perform tests
on these ceramic to prove their usability.
An obvious first criterion which must be met for this materials containment problem is the melting
point of the ceramic. This however is not as easy as it seems, as the regolith electrolyte contains
a mixture of many oxides, which results in the possibility of forming any number of low melting
point eutectics with the containment material, a situation which would result in the rapid failure of
the crucible. The determination of the low melting eutectics of the entire five component system
would require analysis of a very complex phase diagram, which is not available. The data shown
in Table 3, however, can be used to give a rough picture of the possible results.
Table 3: Binary eutectics of various components of the potential regolith-crucible system (°C)8
Oxide Containment Materials
JSC-1A
Constituents
Al2O3
Al2O3
X
CaO
1360
MgO
1995
SiO2
1546
ZrO2
1710
CaO
1360
X
2300
1436
2140
FeO
1310
X
X
X
X
MgO
1995
2300
X
1543
2113
SiO2
1546
1436
1543
X
1675
TiO2
1750
1600
1540
1750
The only common oxide material which does not form binary eutectics below the process
temperature with any of the regolith components is zirconium oxide. However, zirconium oxide
undergoes a phase transformation at 1170 °C which necessitates the inclusion of a stabilizing
additive (typically magnesium, yttrium, or calcium oxides); this potentially impacts the corrosion
performance of the material.
Aluminum oxide is a standard material for high temperature use. In a pure form, it is inert to many
materials at high temperature and is therefore used in applications requiring high temperature
1
corrosion resistance. It is also relatively inexpensive, which makes it an attractive material for
consideration. A second ceramic material which has been known to exhibit good corrosion
resistance, especially towards silicate melts, is aluminum oxide containing a high percentage of
chromium oxide. These materials are used extensively in the glass industry for containment of a
variety of glasses and have been shown to exhibit superior corrosion resistance to glass
compositions similar to JSC-1A near the same processing temperatures9,10. Chrome oxide
spinels are also used for lining waste incinerators as they have been shown to exhibit good
corrosion resistance to the multi component systems developed in these environments11,12.
An unfortunate situation with corrosion is that while equilibrium data, such as phase diagrams,
give an idea of the performance of a material in a particular environment, they do not necessarily
predict the behavior of a real system. A eutectic formation may be arrested by saturation of
solute, especially when kinetic effects of dissolution, which may prevent corrosion because of
limited agitation, are considered. Alternatively, microstructural or processing influences in a
material may be the dominant factor in affecting performance. Grain boundary attack or defect
related failure is difficult to predict. It is necessary therefore to run tests on ceramics under
simulated process conditions to determine conclusively whether they experience corrosion or
withstand exposure to the process environment. Based on their ready availability and low cost,
aluminum, magnesium and zirconium oxides were chosen to be evaluated for performance. While
many other oxide ceramics are possible, including combinations of several oxides, it was decided
to focus on commercially available products rather than custom made components from more
expensive materials. Chrome oxide containing refractories were not investigated because of their
tendency to form volatile carcinogenic hexavalent chromium compounds.
3.1.2 Non-Oxide Ceramics
Non-oxide ceramics for high temperature use are typically carbides, nitrides and borides. These
ceramics are generally known to possess good corrosion resistance, as well as extreme heat
13
resistance, some up to temperatures of well over 3000 ºC . Ceramics with temperature capability
in this range, however, are still under development. The most commonly used non-oxides include
silicon carbide, silicon nitride and boron nitride. These materials find use in applications requiring
high hardness, temperature and corrosion resistance. The major potential advantage non-oxide
ceramics provide for this application is that many are not wetted by molten metals and have
better resistance to wetting by glasses than do oxides14,15. This characteristic is advantageous
because it limits the interactions which take place between the substrate and the corroding
medium. These ceramics, especially boron nitride often have a high degree of corrosion
resistance to both oxide and metallic melts. Additionally, hot pressed BN has high thermal
conductivity and low thermal expansion; giving it high resistance to thermal shock, an important
consideration for the device created in this project.
2
The main drawback to the use of non-oxides in this application is the fact that all of the commonly
available non-oxide ceramics have limited use in oxidizing environments. The maximum
recommended temperature for many non-oxides is under 1000 ºC, far below the 1600 ºC
processing temperature required for molten regolith electrolysis. However, some non-oxide
ceramics have the ability to form protective coatings against oxidation depending on the
environment. Ceramics such as SiC, MoSi 2, and Si3N4 are able to form silicon dioxide glass
coatings which prevent diffusion of oxygen into the bulk of the material, thus preventing
oxidation16. The formation of a protective coating however, may be adversely influenced by the
interaction with the oxide melt, which has the potential to dissolve any glassy coatings forming on
the surface of a ceramic. Additionally limiting the use of non-oxide ceramics is the fact that they
are expensive, and therefore are not manufactured in a wide range of products. Lack of
availability is a major obstacle to the use of many advanced non-oxides which at first glance
might be potential candidates. The two non-oxide ceramics most commonly used in the
temperature range desired are silicon carbide and boron nitride. The melting/sublimation points of
these materials are around 2700 ºC and they are readily available from several manufacturers,
therefore investigation into these ceramics was considered before other non-oxides.
3.2 Experimental Procedure
3.2.1 Stage One
The first stage of the compatibility studies consisted of spot contact testing of various substrates
with JSC-1A lunar regolith simulant. Samples of a range of possible containment materials were
procured and are listed in the table below.
Table 4: Ceramics evaluated in preliminary compatibility trials.
Sample#
1
2
3
4
5
6
7
8
9
11
12
13
14
Comments
Material Description
Alumina
Alumina
Alumina
Alumina
Zirconia
Zirconia
Zirconia
Magnesia
Magnesia
Boron Nitride
Boron Nitride
Silicon Carbide
Silicon Carbide
99.6%
99.8%
99.8%
99.8%
MgO stabilized (open porosity)
CaO stabilized
10.5% Yttria stabilized
Refractory (open porosity)
4.5% closed porosity
Hot Pressed
Hot Pressed, 5% CaO Binder
Alpha phase, pressureless sintered
P. A. D. (hot pressed)
3
These materials were loaded with 1 gram pellets of pressed powder and loaded into a resistance
heated furnace for melting of the powder into a glass similar to the starting composition inside a
MRE cell. The ceramics were heated to a temperature of 1600 °C and held for various times to
evaluate corrosion progression over time. Time at temperature varied from zero to 12 hours. The
ramp rate for heating and cooling was set to 5 °C per minute to minimize the chance of thermal
shock damaging either the test substrates or the containment vessel. Tests were run in both air
and in argon to evaluate the effect of oxygen on corrosion of the ceramics and oxidation of the
non-oxide ceramics.
3.2.2 Stage Two
The second stage of corrosion testing was focused on evaluating the effect of iron oxide on the
corrosion performance of alumina. Electrolysis experiments conducted at MIT on JSC-1A or other
similar melts were seen to have differing performance based on the oxidation state of iron in the
melt. On the moon only iron (II) oxide exists, however because this is an expensive component,
trials were run on melts containing iron (III) to explore the chemistry of electrolysis with lower
costs. Several experiments were conducted on alumina sample #1 from Table 4 because this
was found to be the best performer in the initial investigation. For these experiments 30cc
alumina crucibles were loaded with 35 grams of mixed powders, (Sigma Aldrich), and were
heated in a resistance furnace to 1600 °C and held for 5 hours. These samples were cross
sectioned after exposure to examine the depth of penetration. The compositions used are shown
in Table 5. Tests were run in argon for all samples and also in air for samples with iron (III) oxide
and no iron. Tests with iron (II) were not run in air because this would have simply been oxidized
further to iron (III) resulting in a duplicate of the other condition.
Table 5: Corrosion experimental melt compositions.
Compound
JSC-1A No Iron
Oxides
Comp 1
Comp 2
Comp 3
Comp 4
FeO
0.00
5.00
10.00
0.00
0.00
Fe2O3
0.00
0.00
0.00
5.00
10.00
SiO2
58.07
55.17
52.26
55.17
52.26
Al2O3
18.28
17.37
16.45
17.37
16.45
MgO
10.97
10.42
9.87
10.42
9.87
CaO
12.68
12.05
11.41
12.05
11.41
Total
100
100
100
100
100
3.3 Results
The following series of photos and micrographs give a qualitative picture of the performance of
various ceramic materials exposed to lunar simulant JSC-1A.
4
Figure 2: Corrosion specimen after removal from furnace, numbers correspond to Table 4. These
samples were held for six hours in air. (Scale reference: side lengths of samples 11, 12 are 2.54 x 10-2
m (one inch)).
Figure 3: Specimen after heating to temperature in argon and immediate cooling; no hold was
performed. (Scale reference: side lengths of samples 11, 12 are 2.54 x 10-2 m (one inch)).
5
Figure 4: Alumina sample #1 after an eight hour hold in air. Inset shows the untested structure of
the ceramic for comparison. No penetration by the regolith simulant is apparent.
Figure 5: Boron nitride sample # 11 from Figure 3. Left shows interface reaction product, right
shows BSE SEM picture of same region; labeled areas one and two are identified by EDS in Table 6.
6
Table 6: Standardless EDS results for corrosion product shown in Figure 5.
Element
C
N
O
Na
Mg
Al
Si
Ca
Cr
Mn
Fe
Total
Scan 1
wt %
at %
6.75
11.47
--40.87
52.1
0.85
0.76
4.55
3.81
13.21
9.98
22.84
16.59
9.03
4.59
----1.9
0.69
100
100
Scan 2
wt %
at %
5.43
15.24
5.2
12.51
6.09
12.84
--1.52
2.11
3.22
4.02
9.28
11.13
1.21
1.02
1.08
0.7
0.19
0.12
66.78
40.31
100
100
Figure 6: Cross sections of experimental specimen. Numbers correspond to conditions in
Table 7.
Table 7: Conditions of experiment from Figure 6.
Sample
Number
Composition Number
(Table 5)
Atmosphere
1
2
3
4
5
6
7
8
1
2
3
1
2
3
4
5
Air
Air
Air
Argon
Argon
Argon
Argon
Argon
7
3.4 Discussion
3.4.1 Stage One
Oxide Ceramic Performance
It was found that different alumina materials have different degrees of corrosion resistance. Most
likely this is the effect of the various processing methods and impurities in the material. The #1
alumina sample has the highest corrosion resistance. This material was found to resist both
corrosion and softening at 1600 °C for holding times for up to 12 hours, long enough for current
scale electrolysis experiments. The other alumina samples tested did not exhibit nearly the
resistance to interaction of sample #1. Counter to intuition, it was observed that the lower purity
material was the best performer, with lower melt penetration and deformation of the samples. In
the higher purity samples penetration of the regolith simulant melt into interconnected porosity
within the alumina was considerable. However, in the lower purity sample no penetration was
observed. The lack of penetration was likely due to a higher obtainable density through the use of
sintering agents. For the alumina material SEM analysis suggests no other failure mechanism
than that the ceramic simply dissolves from the interface into the regolith melt. Upon cooling the
glassy melt solidified into a crystalline high alumina phase near the interface and a glassy silica
rich phase.
Other oxide ceramics tested exhibited less than satisfactory performance. All zirconia samples
had compatibility issues which prevented them from being assessed as possible containment
materials. Both the calcium oxide and yttrium oxide stabilized zirconia samples exhibited
intergranular corrosion of the surface exposed to the regolith melt. The yttrium stabilized zirconia
performed better than the calcium stabilized zirconia but still suffered from the formation of a
weak, porous layer, in addition to exhibiting poor thermal shock resistance. The magnesia
stabilized zirconia sample had such a high degree of porosity that it absorbed the entire mass of
simulant placed upon it (Figure 2, sample #5).
Both magnesia ceramics tested were found to be inadequate. The sample which was highly
porous again absorbed the simulant (Figure 2 sample #8). The other magnesia sample had a low
degree of porosity, but was heavily corroded by the regolith at the air-melt interface eliminating it
as a candidate. This is not a surprising result as a basic crucible material is not typically used with
acidic, high silica, melts.
Based upon the comparative performance, the best oxide tested is the alumina sample # 1. It is
the best choice for molten regolith containment for electrolysis experiments. However, given that
8
it does dissolve in the melt it would not be suitable for long term containment unless an allowance
was built into the container for corrosion.
Non Oxide Performance
The performance of non-oxide ceramics in contact with a regolith melt was considered for their
use as both electrolysis containers as well as for material handling components. Two failure
mechanisms are possible for these materials, oxidation and corrosion by the melt. Because the
maximum manufacturer listed temperature for the BN samples tested was only 850 °C there was
concern that oxidation would be considerable at higher temperatures. However, results from
testing show that even after exposure to static air at 1600 °C for 6 hours, the oxidation depth was
only 4 x10-3 m. As can be seen in Figure 3 the regolith solidified in argon has a high contact
angle, meaning that the melt did not wet the substrate. The contact angles in air (Figure 2) are
somewhat lower because of the formation of boron oxide on the surface of the substrate.
Typically high contact angles are indicative of little interaction. Figure 5, however, shows that at
the interface of the boron nitride and the JSC-1A melt small precipitates of a corrosion product
are visible. Based upon EDS results, these precipitates are a reduced metallic component of the
Si and Fe oxides in the regolith. Thermodynamic evaluation of the following equations show that
a reaction of this type is expected for iron oxide, but is unfavorable for silicon implying a more
complicated reaction path than simple reduction17. In the equations, the subscript (a) implies that
the component is incorporated into a glass.
BN( s )
Fe2O3
2Fe(l )
4BN( s ) 3SiO2( a )
B2O3( a )
N2( g )
3Si(l ) 2B2O3( a ) 2 N2( g )
(1)
(2)
The reactions shown above took place in limited locations rather than across the whole surface,
thus a pitting phenomenon is observed. The corrosion product penetrated only minimally
however. It is likely that the maximum depth of penetration is related to the amount of reducible
component in the melt. Once all iron oxide is reduced, the reduction reaction should cease and
corrosion should be limited to simply dissolution, which is slower than the observed reaction.
Experiments to validate this proposal were not undertaken, so a definitive behavior cannot be
described.
The results obtained imply that BN is a feasible containment material for use in thermal shock
susceptible applications, provided exposure to high partial pressures of oxygen is limited. Given
its low wettability, it would be especially suitable for applications in which a component is desired
to drain completely or remain free of frozen metal or glass after liquid transfer to another
container.
9
Silicon carbide is known to have good high temperature oxidation resistance because of the
formation of a silicon dioxide layer. However, in contact with the melt in air it was found that a fair
amount of oxidation took place. In Figure 2 a sample of silicon carbide is shown which is coated
with a white substance. This substance is foamed regolith, which foams due to the formation of
CO bubbles. This foam layer was not observed in the samples that were tested in an inert
atmosphere. Results which are not pictured showed that SiC does react with the melt to form a
SiC + Fe3C (as suggested by EDS analysis). There is also an interfacial reaction layer high in
carbon. This is consistent with literature reports of SiC corrosion by iron18. Again pitting corrosion
is exhibited in SiC and could be problematic if the material is exposed to the regolith melt for
extended time. Corrosion penetration was more significant in SiC than in BN.
It is apparent from comparing the contact angles of solidified beads (Figure 3) that SiC is not
wetted by the melt to the degree of the oxide ceramics; however it is wetted more than the BN
samples. Because of the attack exhibited by contact with molten regolith and the concern for
oxidation this material is probably not worth using as a long term container.
For long term containment in a MRE cell neither non-oxide ceramic is a permanent solution. Both
because of red-ox reactions between the melt and the containers and because of atmospheric
oxidation instability these containers would need to have some corrosion allowance built in to
prevent failure. The high purity grades tested are expensive, suggesting that they would not make
a cost effective containment strategy. While larger components of lower grade material could be
used with the acceptance of some corrosion, because a reasonable material, (alumina), was
found for containment on the scale of 10 hours it was decided not to evaluate their performance
further.
Though it was determined that boron nitride and silicon carbide are not good solutions for long
term containment, the degree of degradation experienced in these trials suggested that for short
term exposure they could prove adequate. Because of their thermal shock resistance and low
wetting it was decided to attempt using these materials for fluid handling components for the main
objective of the project, molten fluid removal. Additionally supporting their use for this purpose
was the fact that both non oxide ceramics degraded as a result of interaction with iron oxide
rather than iron itself. The material to be removed from the cell is likely to be absent of iron
oxides, the entire iron oxide content having been reduced to the metal. Therefore reaction with
the depleted regolith melt should be less severe. While it is known that iron reacts with SiC, it is
still a widely used product for ferrous melting and handling purposes. Additionally, it has been
shown that for melts containing silicon above a temperature dependent critical point reaction
10
18
proceeds much more slowly . Since the MRE process produces not pure iron, (as was the initial
product in these experiments), but a silicon rich Fe-Si alloy, the metal to be removed was not
expected to have a great deal of interaction with silicon carbide handling components.
3.4.2 Stage Two
Based upon results seen at MIT during electrolysis runs, it was expected that the addition of iron
oxide to a silicate melt would increase the amount of corrosion of the alumina crucibles. Iron
oxide is a network modifier and therefore reduces the viscosity of silicate melts. A reduced
viscosity allows faster transport of dissolved species and therefore ignoring any other reactions
taking place, the addition of iron should allow faster dissolution of the crucibles.
However, the results of corrosion experimentation with the sample #1 alumina indicate that rather
than iron, the greater influence on corrosion was the atmosphere above the melt. Figure 6 shows
that penetration at the melt-atmosphere interface was dramatically higher for those samples run
in argon than the samples run in air. Additionally apparent from the figure is that the melts
containing iron (III) oxide appear dark brown in the air melted specimen (2 and 3) and bluish in
the argon melted samples (5 and 6). The influence of iron (III) in silicates is to turn them green or
brown while iron (II) makes the glass appear blue. The colors of the specimen are indication that
the atmosphere in the argon trials was actually reducing rather than inert, changing the iron (III) to
iron (II). This can be ascribed to the fact that the furnace used had graphite heating elements.
Without a means of sealing the samples from the furnace atmosphere it was decided not to
attempt to re-run these experiments in a completely inert environment.
Results might seem to indicate that iron (II) is the more reactive component. However,
observation of the iron devoid samples shows that when run in air very little corrosion takes
place, while under a reducing atmosphere considerable corrosion occurs, perforating the crucible
wall in one case. Therefore it appears that rather than iron oxidation state the greater influence on
corrosion of the alumina tested is the atmosphere above the melt.
3.5 Conclusions
It was determined that the best performer for long term contact with a JSC-1A type melt was a
99.6% alumina. This material exhibited less corrosion than did higher purity alumina containers
because it had less porosity. In the higher purity materials the porosity was infiltrated by the
silicate melt resulting in deformation of the sample and increased interfacial interaction. This
material performs better under air than under a reducing environment. It is suggested that the
influence of iron oxidation state is not a significant factor relative to that of the oxidizing nature of
atmosphere above the melt. Atmospheric conditions may be the true reason for the difference in
11
results seen in crucible performance during electrolysis experiments. Further testing in a sealed
inert atmosphere is required for confirmation of the proposed relationship.
Given its performance in the corrosion tests described, this material was used as a container for
experimental withdrawal trials containing both ferrosilicon and depleted regolith in a dual layered
melt. In this configuration this alumina was shown to withstand contact for up to 9 hours with a
475 cc melt, half the volume of which was metallic and half oxide.
Corrosion testing also indicated that both boron nitride and silicon carbide exhibited corrosion
resistance to the melt sufficient for short term contact. The main driver for corrosion was reaction
with iron oxide in which the non-oxide substrate reduced the iron from the melt. These
components were also wetted less than the oxide materials, with boron nitride exhibiting the least
wetting
12
Section 4.
Design of the Experimental Apparatus
As the feasibility of the electrochemical process of molten regolith electrolysis has been proved,
the primary purpose of this work was to identify a means of removing molten materials from an
electrolysis cell. Development of such a method is critical to creating a prototype device for
exploring issues with the full process, charging, electrolysis and removal, on a practical scale.
4.1 Design Constrains of the Fluid Removal Device
The design constrains of the molten material removal device are listed below.





Capable of removal of molten metals similar to the products of the molten regolith
electrolysis cell
Capable of removal of molten oxide „electrolyte‟ similar to the depleted lunar regolith
Functionality compatible with oxygen capture from the electrolysis cell
Operational capability at 1600 °C
Limit or eliminate consumable materials
These demanding requirements called for an unusual approach to design considering not only
the mechanical function of the device but also the materials selection for operation in a highly
reactive environment.
4.2 Current Methods for Withdrawal of Molten Fluids and
Method Selection
A simplified geometry of the proposed regolith electrolysis cell is represented in Figure 7.
Because of the identified difficulty of sustainable containment using a hot walled cell the
configuration shows what is known as a cold walled reactor. Such systems are used when the
material to be contained is too corrosive for contact with a container or cannot tolerate
contamination. One example of cold walled containment in industrial practice is titanium casting in
which titanium is melted using induction heating in a skull melting furnace. Skull melting is also
used for melting oxides, commonly cubic zirconia or other technical ceramics. Of greatest
relevance to this project is the cold walled process used in the typical Hall-Heroult cell for
aluminum production. In the Hall-Heroult cell large scale cold walled melting is used to contain
the cryolite electrolyte which is used for electrowinning of aluminum from dissolved bauxite.
1
Figure 7: Cold walled cell schematic2.
In each of these processes a different method of removing the material in the furnace is
employed. The simplest conceivable means of removing molten material from a furnace is to tip
the furnace over and allow gravity to dump the fluid out. For some cold walled processes this is
done. Titanium is cast by simply tipping the crucible and pouring the melt out into the mold.
However for the MRE application this is not an applicable solution. First, it is desired to maintain
accurate control over the level of fluid in the cell, both of the ferrosilicon produced and of the
electrolyte. Pouring of a two layered melt would be difficult to control on a repeatable basis.
Additionally pouring is not a reliable means because pouring the melt over the lip of the container
requires that the lip be capable of handling contact with high temperature fluid at roughly 1600
°C. This vessel also must seal air tight for containment of oxygen against the ultrahigh vacuum on
the lunar surface. Sealing against high vacuum and handling high temperature contact would
require a system beyond the concepts conceivable by the author or others involved in the project.
Finally, the crust which would freeze over the top of the furnace presents another difficulty in
simply pouring material from the cell.
Another gravity operated mechanism for withdrawal uses a hole in the bottom of the container
which could be opened for dumping of fluid (Figure 8). Plugging this hole would require either use
of an actuated mechanism or allowing the material to freeze in the hole. Industrial methods for
plugging a tap hole include use of a tapered ceramic plug or a slide gate which is opened for
allowing fluid flow.
1
Plunger
Motion
Tap Hole
Slide
Gate
Motion
Figure 8: Mechanisms for closing gravity operated fluid removal methods. Left: slide gate, Center:
tapered plunger, Right: Ceramic plug in burned hole.
Both of these methods are employed widely for use with hot walled cells in which the fluid is in
contact with the wall of the crucible. In a cold walled cell the crust surrounding the molten material
would need to be broken in order to access the fluid, necessitating a means of breaking the crust
in addition to plugging the hole. In both of these configurations a good mechanical seal is
required to prevent leaking from the cell. Without an operator present, as is typical in industry, it
would be difficult to enable use of these methods for controlling fluid containment. Also, as
discussed earlier, since there is no material with good corrosion resistance to both the metallic
and oxide products for extended periods of time, selecting a material for use as a plug or slide
gate is difficult. Industrially this problem is solved by simply considering the critical components
as consumables. Crucibles and ladles are relined as they inevitably wear. Slide gates and tap
hole plugs are regularly replaced as they corrode or are clogged. This luxury of using
consumables does not exist on the moon, thus eliminating the tapping and plugging method as a
possibility.
Another means of removing liquids from a container is through the use of a siphon type device.
Such a device is depicted in Figure 9.
2
P1
P2>P1
Figure 9: Counter gravity materials removal. Pressure differential can be created either by applying
pressure on the melt or by pulling a vacuum over the removal tube.
This method uses a difference in fluid pressure between the level of a tube inlet and outlet to
transport material up the tube and out of the reactor. Such a method is used industrially to
produce castings in what is called counter gravity casting. In this process a tube is dipped below
the surface of a melt and vacuum is applied over a mold to pull metal up into the mold in a slow
controlled manner, reducing turbulence and air entrapment and preventing damage to the mold.
The process can be used for a range of metals including steels and superalloys which are cast
near 1600 °C as would be done in a MRE cell. The tube which is used for drawing metal is
typically an integral part of a ceramic investment mold and is thus used once and destroyed. In an
alternate variant of this process, pressure can be applied over a sealed ladle into which a tube
has been inserted. In this case the tube can last for many shots of metal, but after some time it is
still destroyed19.
A siphon method is used for removal of metal from the Hall-Heroult cell. Typically, a steel tube is
employed for pulling aluminum from the cell near a temperature of 960 °C20. This method allows
the tube to be punched through the frozen cryolite crust at the top of the cell and below the
surface of the melt into the aluminum layer at the bottom of the cell. This allows accurate removal
of aluminum, giving process engineers control of the level of fluid in the reactor. It also removes
concern about function and reliability of a tap apparatus. This method can be used for small scale
fluid transfer, or as demonstrated by the aluminum industry, can be scaled to work with a large
cell.
3
While it is conceivable that a gravity operated mechanism could be made possible by
containment of the entire electrolysis and removal system inside a pressure tight vessel this
would call for a large, heavy and expensive system. In addition to requiring a large external
pressure vessel, until the cold walled cell configuration is developed any work on development of
the electrolysis process will be done in an externally heated furnace. Removing material from this
setup would be quite difficult given the need to accommodate the furnace geometry and
operation. Because of its flexibility to work on a range of scales and in a limited access
environment the siphon, or pressure differential method, was chosen as the best means for
removing fluids from a MRE cell.
4.3 Description of the Experimental Apparatus
4.3.1 Basic Principles
The critical components of a pressure differential withdrawal device are:




A container for holding molten fluid during melting
A container for receiving molten fluid after removal from the primary container
A tube for connecting the two vessels
A method of achieving a pressure differential between the two vessels
The basic operation of the device consists of creating a lower pressure in the receiving vessel
than in the primary container and allowing the difference in pressure to push fluid through the
connecting tube into the receiving container. When sufficient fluid is transferred the pressure
differential is removed and fluid transport stops. Possible means of achieving a pressure
differential are either pulling a vacuum on the receiving container and allowing atmospheric
pressure to be the pressurizing means, or applying a positive pressure over the primary container
giving it a higher total pressure than the receiving container at atmospheric pressure. The merits
of the two options are discussed further in the following section, but for this experiment the
method selected was the use of partial vacuum over the receiving container.
4.3.2 Individual Design Considerations
The mechanical design, materials selection and the geometric specifications of the system were
done by approaching each individual component in series using the limitations of the previous
component as a starting point for the design of the next. Some dimensions, which because of the
flexibility of the new design were initially arbitrarily selected, became design limiting as
modifications to the hardware were made later in the project. The actual dimensions of the
system are specific to the configuration used based on the financial constraints of the project.
Were a new design to be implemented, the dimensions of many components could be changed
allowing a more efficient design. Additionally affecting the design of this system is the fact that no
electrolysis was performed during any of the withdrawal experiments. As of this writing, the
electrolysis hardware space requirements are still undefined. Given the flexibility of both the
4
electrolysis and pressure differential withdrawal hardware, it was decided to proceed with a
design that ignored the need for inclusion of the electrical connections and electrodes in the cell.
As the requirements of both components of the MRE process were determined, adjustments
could be made to the design to allow incorporation of both the electrolysis and withdrawal
hardware. This section describes the final design considerations without discussion of any
intermediate steps during revision based on experimentation.
Melting Furnace
The starting point for the design and the most expensive component of the system was the
furnace used for melting the lunar regolith simulant and ferrosilicon alloy to be withdrawn. Given
the roughly $40,000 price tag for a new furnace that was to have only a short period of use, the
decision was made to instead refurbish an existing furnace and modify it for use with the
materials withdrawal device. A furnace available at Marshall Space Flight Center met the
requirement of 1600 °C continuous operation. This furnace was selected as the choice for this
project. Manufactured by Deltech Inc, Denver Co., the furnace insulation was replaced and new
top insulation was installed with a 1.40 x10-1 m (5.5 in) bore hole to allow access to the inside of
the furnace during experiments. A round plug was fabricated to fit in the bore hole and close the
furnace allowing it to reach high temperature. The furnace has a 2.54 x10-1 m (10 in) bore and an
approximately 2.80 x10-1 m (11 in) tall hot zone. The plug itself was bored to 1.14 x10-1 m (4.5 in)
in diameter to allow an alumina furnace tube to drop through the plug and furnace borehole to
contain the crucible in event of any leakage.
Crucible
The next component required for the device was a crucible for holding the molten material to be
withdrawn. As determined in the corrosion testing, there is not a suitable material for indefinite
resistance to attack from all three elements present in the MRE cell environment. The best
performer, 99.6% alumina, has sufficient resistance to lunar regolith simulant and ferrosilicon to
contain the melt for at least 15 hours at temperature, long enough for the withdrawal experiments.
The size of the crucible was chosen to be a cylindrical 500 cc crucible, mainly because it is a
standard size and allows the ability to nest this crucible inside a larger 750 cc crucible for
protection against breakage of the primary containment.
Mold
While it would be convenient to have a permanent mold fabricated of copper or cast iron, low cost
and the potential for redesign made an expendable sand mold the best choice. Resin bonded
silica sand, held with sodium silicate resin, was used as a molding material for creating molds.
Advantages of this choice included the low cost of the mold and eliminating a mold coating which
would have been required with a permanent mold. The main drawbacks to using sand molds
were that it required careful handling to prevent getting sand on sealing surfaces and that it was
5
difficult to attach sensors to a sand surface. With the exception of the volume, the mold
dimensions were rather arbitrary, simply allowing 2.54 x10-2 m (1 in) of wall thickness to give a
strong mold and prevent breakage during filling (see Figure 10). The volume of the mold was
determined based upon the volume of the crucible and the volume of the withdrawal tube used to
transfer material. It will therefore be addressed in the next section.
Figure 10: General dimensions (mm) of the sand mold used to receive the withdrawn melt. The inside
volume up to the ledge is 283 ml.
Withdrawal Tube
The tube for transfer of molten material, or withdrawal tube, was the most critical component of
the system. Several different withdrawal tubes were tested with varying geometries and materials
to examine the performance differences.
Evaluation of ceramics for long term containment showed that there is no material with sufficient
stability in molten regolith for time periods long enough to allow continuous immersion of a
withdrawal tube in the melt during electrolysis. This requires that the tube be introduced into the
melt just prior to withdrawal to avoid corrosion. Using this method reintroduces the possibility of
using oxidation prone non oxide ceramics or even melt-soluble refractory metals given that only
brief exposure to the electrolysis products is experienced during withdrawal.
In order of importance, the criteria used in the selection of a material for the withdrawal tube
were: operational capability at 1600 °C, thermal shock resistance, melt wetting behavior, and
oxidation resistance. Simply imposing the temperature limit eliminates most metals and even
6
many ceramics. Thermal shock requirements and the high degree of wetting by glass make an
oxide ceramic an unattractive choice, and little oxidation resistance and availability renders use of
non-oxide ceramics challenging. Because the constraints on the material were conflicting and
operational parameters not firmly dictated it was decided to try using materials from each class to
evaluate both material and operating procedure changes. Molybdenum was chosen as the
cheapest available refractory metal, silica and alumina were oxide ceramics tested, and silicon
carbide and boron nitride were non oxide ceramics used for their shock and wetting resistance.
This approach allowed the performance of each material to be empirically evaluated rather than
trying to select based on strictly theoretical considerations.
Using a removable rather than fixed tube required a means of dropping the tube into the melt and
removing it immediately after withdrawal. This was accomplished with the use of a ball screw
driven linear actuator. With a 9.14 x10-1 m (3 ft) total travel the actuator was selected to give
enough clearance over the furnace to allow complete removal of a 6.10 x10-1 m (2 ft) long
withdrawal tube. Ball screw drive also ensured accurate, repeatable control of position and
resistance to short term exposure to high temperature gas above the furnace during withdrawal.
Additionally, attaching the tube to an actuator allowed adjustment of the height of the tube inlet,
meaning that control over withdrawal of a multilayered melt was possible.
Two difficulties are created by using a removable rather than fixed tube. First, introducing the
tube cold into the furnace presents thermal shock concern; second, using a cold tube for
withdrawal causes chilling and potential freezing of the melt. Thermal shock was prevented by
slow travel of the tube into and out of the furnace while the chilling issue was addressed by
preheating, as will be discussed in the experimental section.
The geometrical constraints of the withdrawal tube were given by the furnace dimensions and by
the flow and thermal properties of the fluids to be transported out of the furnace, (MRE cell). The
first aspect to be considered was the length of the tube. The location of the hot zone of the
furnace, the thickness of the furnace insulation and the space required for the heating elements
mandated that the tube be at least 0.45 m (18 in) long. This minimum length was used, to enable
removal of the maximum amount of material from the crucible and limit cooling during withdrawal.
With the minimum length of the tube fixed, the inner diameter of the tube could be specified
based on the rheology of the fluid. Equations 3 and 4 describe the velocity of fluid moving in a
vertical tube.
vz
p
l
g
7
R2 r 2
4
(3)
v
R2
g
8
p
l
where v z is the fluid velocity up the tube, v is the average velocity,
differential from the top to bottom of the tube,
l is the tube length,
(4)
p is the pressure
is the fluid density and
the fluid viscosity.
As shown in the equations, the viscosity of the withdrawn fluid was of critical consideration to
determining the diameter of the tube. Figures 11 and 12 show the viscosity temperature curves
for the fluids to be withdrawn. At the time of design the viscosity of the molten regolith was not
know so a design estimate for the viscosity of the mixture at 1600 °C was approximately 0.63
Pa·s plus or minus an order of magnitude based on measurements of similar silicate melts21.
Later work on identifying fluid properties revealed that at the withdrawal temperature 0.63 Pa·s
was a good estimate.
Figure 11: Temperature-viscosity curves for compositions similar to a depleted regolith melt22.
8
is
Viscosity (Pa-s))
2.5
2.0
1.5
1.0
0.5
0.0
1350
1450
1550
Temperature (C)
1650
1750
Figure 12: Temperature-viscosity curve estimate for 60wt% Si 40% Fe. Viscosity obtained by mol
fraction average of Si and Fe metals. Viscosities below melting points obtained by linear
extrapolation 23,24.
By adjusting variables within equations 3 and 4 an appropriate inner diameter for the tube was set
at 1.91 x10-2 m (0.75 in). With this diameter, the withdrawal volume of 300 cc could be removed
at a manageable 0.25 m/s, (71cc/s), in approximately 4.2 seconds using the pressures noted in
Table 8. For these calculations the Reynolds‟s numbers for flow are 10.1 and 20.1 respectively,
well within the laminar flow regime described by the following equations.
Table 8: Pressure differentials for constant temperature, steady state withdrawal in earth gravity.
Material
60wt%Si 40%Fe
Depleted Regolith
Viscosity at
1600 °C
Pa*s
1.4
0.63
Density at
1600 °C25
kg/m^3
3200
2650
Pressure Differential for
0.25 m/s avg. velocity
Pa
PSI
28,500
4.14
18,250
2.65
In addition to steady state flow, the equations used to predict fluid flow also assume a constant
viscosity within the fluid. Unfortunately, for this case neither of these assumptions is accurate.
The viscosity of oxide glasses has an inverse exponential relationship with temperature so if a
glass cools during withdrawal through a cold tube its flow will dramatically slow. Ignoring the
chilling effects of the tube on the melt as it is withdrawn therefore predicts much easier flow than
would be experienced in a real situation in which a cold tube is introduced into the crucible, (or
cold walled MRE cell), just prior to withdrawal. Further complicating prediction of flow behavior in
the tube is that the situation is not steady state. Equation (3 assumes that the tube is completely
full of fluid and that the pressure drop across the tube is fixed. In the case of the real system,
pressure drop is dynamic and fluid height in the tube increases until it overflows into the mold.
The transient nature of both thermal and mass transport during the period of interest makes it
9
difficult to use analytical evaluation to make definite quantitative predictions of fluid behavior
during withdrawal. Because the two assumptions lead to opposite simplifications of the system,
(ignoring heat transport suggests easier removal while assuming steady state flow gives a
conservative estimate of requirements), they offset each other to some degree. The steady state
calculation was therefore used as a first approximation for the selection of a tube diameter and
1.91 x10-2 m (0.75 inch) was used.
Figure 13: Boron nitride coated molybdenum tube. Notch for directing outlet flow on right. Uncoated
portion allows better gluing.
The tube diameter used allowed for both a reasonable estimated withdrawal time of the molten
oxide and gave a standard size for which materials would be easy to procure. Using the smallest
diameter tube was desirable because it minimized the internal volume of the tube. A smaller tube
volume meant that for a given volume of melt in the crucible more material could be withdrawn
2
-1
(removable volume = crucible volume – (tube volume + safety volume )). Using a tube 4.57 x10
m (18 in) long x 1.91 x10-2 m (0.75 in) ID the maximum possible withdrawal volume was
approximately 300 cc. The mold was designed to have this volume. The outer diameter of the
tube was arbitrary and varied between experiments but the tube wall thickness was kept small in
an effort to minimize the thermal mass of the tube to prevent chilling of the melt during
withdrawal. A molybdenum tube used for this experiment is shown in Figure 13.
Pressure Differential Components
The primary consideration in addressing whether to use a pressurized system or a vacuum
system to run the withdrawal was the issue of sealing the pressure differential. To create a
pressure differential between the receiving mold and the melt contained in the crucible it was
chosen to use vacuum over the mold leaving a pressure difference between the atmospheric
pressure above the crucible and the reduced pressure over the mold. The choice was made
because it was desired to have a seal which could allow variable positioning of the withdrawal
2
Safety volume prevents splashing of removed fluids as described in experimental section
10
tube. While a fixed sealing method can be envisioned for allowing pressurization of the furnace
chamber (electrolysis cell), allowing vertical translation of the tube during withdrawal requires a
dynamic seal. Contact dynamic seals typically require lubrication and are not meant for high
temperature. The seal for a pressurized system must be against the hot tube, eliminating the
dynamic sealing option unless the dynamic seal is a separate mechanism. Several options for
creating a dynamic seal were considered but difficulty or cost of using these methods eliminated
a pressurized cell design for this study. Concepts for how a dynamic seal could be employed are
discussed in Section Section 6.
The sealing method for vacuuming the mold allows a statically sealed tube to translate in the
crucible during withdrawal. This is because the ambient pressure and cell pressure are the same.
This fundamental difference between the pressurized vs. vacuum method is illustrated in Figure
14.
Dynamic Seal
P< 1ATM
1 ATM
T ~1600 °C
1 ATM
P > 1 ATM
P> 1ATM
Static Seal
1 ATM
1ATM
P< 1 ATM
T ~1600 °C
1 ATM
No Seal
Required
Pressurized Withdrawal
Vacuum Withdrawal
Figure 14: Pressurized withdrawal requires a dynamic high temperature seal against the tube if
translation is to be achieved during withdrawal. Using vacuum withdrawal eliminates the need for a
dynamic seal because the pressure inside and outside the furnace (electrolysis cell) is the same.
11
The required components for a vacuumed system were: a closed vacuum chamber, a vacuum
pump and a connecting line with an appropriate valve system for controlling the pressure. An
aluminum dome for pressure containment was fabricated and a ball valve was used for control of
air flow out of the chamber. This allowed an operator to observe the flow of fluid into the mold
through windows and dynamically adjust pressure in the dome by adjustment of the valve
connected to a vacuum source. A roughing vacuum pump connected to a buffer volume was
used as a vacuum source as there was no need for generation of a high vacuum. Pressure in the
mold chamber was measured with a pressure transducer (Measurement Specialties, US300
series) to obtain the time-pressure curve required for withdrawal. Operation of the vacuum
equipment was manually controlled.
4.4 Evolution of Design
The withdrawal system as described in the previous sections is the final configuration used after a
number of experiments. The evolution of the components and operation of the system took place
over a period of months as experiments showed the shortcomings of the previously used
configuration. The general process was to start with as simple and low cost a solution as possible
and build up more functionality as problems were identified. The general design of the system
was not radically changed, but small modifications were made to allow more control over
experiments and improve performance. Figure 15 provides a comparison between the latest
version and the initial configuration. Significant differences between the design previously
described and the initial design are listed in Table 9.
Table 9: Significant modifications between initial and final withdrawal system configuration
Component
Changes from Initial to Final
Vacuum Dome
Valve
Withdrawal Tube
Safety Wall
Glass dome changed to aluminum dome with ports for sensors
Needle valve changed to ball valve to allow faster actuation
Tube material varied and length reduced
Wall created to protect operator in event of splash and dome failure
12
Figure 15: Initial (left) and final (right) configuration used for the differential pressure withdrawal
device.
4.5 Limitations of Current Design and Scale-up Issues
As discussed in the previous sections, the design of this device was limited by the available
equipment for melting the oxide and ferrosilicon melt. In addition to the dynamic sealing issue, a
vacuum system was also necessitated by the use of a resistance heated furnace rather than a
cold walled cell. A pressurized system would require that either a hot walled pressure vessel be
installed in the furnace to contain the crucible, or that the entire furnace be placed inside a large
cold walled vessel. Neither option was feasible for this project given the limited strength of a
pressure vessel at 1600 °C and the cost of a large vessel for containment of the entire furnace. In
lunar operation, however, the pressure differential device must be of the pressurized cell design.
At first glance this is may not be obvious, but the essential criterion for a vacuum operated device
is that the ambient pressure and cell pressure be the same. Thus, on the lunar surface where no
atmosphere exists no ambient pressurizing medium is available and a vacuum system does not
function. This limits the options to either operating a cell with a dynamic seal as noted in Figure
14, or running a vacuum type withdrawal in a vessel which encloses the entire system, (cell,
vacuumed mold chamber and tube actuating equipment).
13
The size requirements of a cold walled cell are yet unknown. The dimensions of the cell are
dependent on establishing a balance of heat flow into the cell from joule heating during
electrolysis and out of the cell wall by conduction/radiation. Therefore the cell size could
potentially be modified by insulating or cooling the cell to achieve the desired geometry. Given
that the cell size is likely adjustable this suggests that the entire cell could be made to fit inside a
closed pressure vessel of reasonable size. This pressure tight shell would enable capture of
oxygen generated and would also allow a pressurized cell withdrawal method.
Running a pressurized cell withdrawal introduces the complicating difficulty of requiring a high
temperature dynamic seal. The dynamic seal is desirable because it gives flexibility over which
components are withdrawn from a multilayered cell (variable height of withdrawal). If it is
determined after development of the cold walled cell that the withdrawal does not require
flexibility in withdrawal tube location then the dynamic seal can be eliminated.
Design of a dynamic seal was investigated but abandoned as it was decided to focus on proving
the pressure differential concept. Though this is the main limitation of the current design several
ideas on how this could be accomplished are presented in Section Section 6.
It has been demonstrated industrially that the pressure differential concept is usable for removal
of metals at weights up to hundreds of pounds and at temperatures in the range of 1600 °C19.
Thus, scaling of the equipment to operate on a larger cell should not present significant technical
challenge.
4.6 Testing Strategy
The general strategy used to prove the pressure differential withdrawal method for removal of
material from a MRE cell was to start with the simplest withdrawal situation, i.e. low temperature,
non solidifying fluids, and work towards progressively more difficult conditions. With this in mind,
the equipment was initially built to function for the easiest withdrawal case. Experimental
performance indicated where adjustments to the system were necessary to improve its
performance for more demanding applications.
The first procedure considered was withdrawal of water from a crucible. Water has a similar
dynamic viscosity to molten metals; (dynamic viscosity of water 1 x 10-6 m2/s at room
temperature, dynamic viscosity molten iron 7.2 x10-7 m2/s at 1627 °C26). This allowed water to be
used as an analogue for the flow performance of molten metal. As a first look at a more viscous
melt, similar to the depleted regolith glass, corn syrup was also withdrawn.
14
Once the behavior of the system was evaluated using water and corn syrup experiments
progressed towards withdrawing molten metals to evaluate the concern of freezing. Aluminum
was used to enable evaluation of molten metal withdrawal at lower temperature than needed for
withdrawal of molten ferrosilicon. Once aluminum was successfully removed the system was
validated for ferrosilicon withdrawal.
Finally work turned towards removal of molten oxides. The feasibility of removal of molten oxide
using pressure differential was a concern because though the viscosities of the fluids are similar
(0.498 Pa·s for molten iron vs. 0.63- 9.3 Pa·s for depleted regolith) the exponential relationship
between temperature and viscosity for glass meant that any cooling during withdrawal would
dramatically increase viscosity. This aspect was seen as the most difficult hurdle in the validation
of the method. It was broken into three steps, first a modified composition was used that allowed
a similar viscosity to the depleted regolith be obtained at a lower temperature. Then, a depleted
regolith melt was withdrawn at 1600 °C. Finally, both ferrosilicon and depleted regolith were
withdrawn simultaneously, similar to the process to be used for lunar implementation.
The following section describes the procedures used for and the results of the various trials
described above and the modifications that were made to the device as a consequence of the
results obtained.
15
Section 5.
Experimental Withdrawal Procedures and
Results
5.1 Water
5.1.1 Setup and Procedure
The purpose of this test was a first evaluation of the system to see if any results did not match the
expected response. The setup for this experiment was used as a basis for the following
experimental setups performing the withdrawal of metals and oxides. The standard setup used for
all experiments will be described in this section; modifications from the setup described will be
noted in the appropriate section.
Setup
The components required for preparing the furnace for a withdrawal experiment were: an alumina
furnace plug for closing the furnace borehole, an alumina furnace tube, a crucible for holding the
melt (water), a secondary containment crucible used in the event of a crack in the primary
containment, and a pedestal used to raise the crucible in the furnace tube. The furnace tube was
lowered into the furnace through the plug and the pedestal was placed in the tube upon which the
crucibles were set. The crucibles used in these experiments were 99.6% high purity alumina
supplied by AdValue Technology as they were identified in the corrosion study to be the best
performers for long term containment. The primary crucible used was a 500 cc cylindrical crucible
with approximate dimensions of 6.70 x10-1 m (2.625 in) inside diameter x 1.43 x10-1 m (5.625 in)
inside height. This crucible was placed inside a secondary crucible of the same material with a
slightly larger, 750 cc, volume. A layer of alumina sand approximately half an inch thick was used
between the crucibles to provide a stable base. The secondary crucible was used to provide
containment in the event of a cracked crucible, an event which had been seen with regularity in
electrolysis experiments at MIT. The crucibles were placed upon a pillar of alumina refractory
brick which was used to minimize the distance from the crucible to the mold above the furnace,
thus reducing the withdrawal tube length. The appendix provides a diagram of the geometry used
for later experiments; it is similar to that used for this experiment.
Above the furnace the withdrawal tube was glued into the tube mounting flange with Thermeez
7200 high temperature alumina paste glue. For the water experiment the withdrawal tube was
-2
-2
-1
2.54 x10 m (1 in) OD x 1.9 x10 m (0.75 in) ID x 5.59 x10 m (22 in) long and was made of
1
fused silica. A container for receiving the fluid from the furnace was placed on the mold plate and
an attachment was glued to the silica tube to allow water sucked up the tube to run into the
container. A borosilicate glass dome was placed over top of the receiving container and tube.
This dome was sealed to the steel mold plate with a foamed silicon rubber gasket. The only
clamping force used was the weight of the dome on the foamed rubber. As vacuum was applied
atmospheric pressure increased the downward force on the dome providing a better seal.
Withdrawal Procedure
The crucible was filled with approximately 400 cc of water which was dyed in order to improve
visibility in the silica tube. Using the linear actuator, the mold plate was driven downwards,
lowering the withdrawal tube to within 6.4 x10-3 m (0.25 in) of the bottom of the primary crucible.
The vacuum pump was turned on and the buffer tank was evacuated. Once the tank was at full
vacuum, the needle valve on the line connecting the tank to the sealed glass vacuum dome was
slowly opened. As pressure decreased in the dome, the column of water rose in the tube until it
overflowed into the receiving container. Pressure control was maintained by visual observation of
the flow behavior of the water and slow adjustment of the valve. Once the container was full the
needle valve was closed and the chamber was vented allowing the water in the tube to drain back
into the crucible.
5.1.2 Results and Discussion
The basic result was as expected; increasing pressure differential between the atmosphere in the
dome and the ambient atmosphere caused a rise in the water column in the withdrawal tube until
overflow emptied the crucible into the receiving container. However, it was discovered that a
significant complicating factor resulted in unexpected consequences when the withdrawal was not
run with controlled measurement of the volume of water transferred. When the water level in the
crucible was above the inlet of the withdrawal tube water rose and overflowed smoothly from the
tube. As the water was withdrawn from the crucible the height of fluid in the crucible dropped,
eventually reaching the point where the water level was at the same height as the tube inlet,
-3
approximately 6.4 x10 m (0.25 in) above the bottom of the crucible in this case. Experience with
a drinking straw suggested that when this situation arose the tube would simply start to draw air
and the withdrawal process would end. This was not the case. Upon reaching this critical level air
bubbles were ingested into the tube and immediately following was a geyser of water which shot
around the inside of the vacuum dome. The suggested process by which the jet occurs is
illustrated in Figure 16.
1
Figure 16: (Following page) Pressure is reduced inside the dome allowing atmospheric pressure to
push fluid up the withdrawal tube (1). As the mold fills the level of liquid in the crucible drops until it
reaches the inlet of the tube, or the critical height, at this point bubbles are ingested in the tube. (2,3).
Once vacuum is lost the fluid in the tube falls back into the crucible raising the height of liquid, once
again sealing the tube (4). Upon resealing the pressure differential is rapidly re-achieved causing the
uncontrolled rise of liquid up the withdrawal tube and splashing into the vacuum chamber (5).
2
1
2
3
4
3
Figure 16 Continued
5
This result was repeated multiple times and was seen to occur in every case in which the water
level dropped to the critical height. For the case in which water was withdrawn, the worst problem
was a puddle inside the vacuum dome. As this process was to be extended to molten metal
handling, a solution was necessary for preventing the jetting issue for both safety and practical
operation reasons.
5.1.3 Conclusions and Modifications
The critical issue of concern brought to light from these experiments was the relative positions of
the tube and the fluid level in the crucible. As it would be difficult to place a sensor in the furnace
to detect the fluid level it was decided that the best way to approach this issue was from simply
knowing the volumes of the components, (crucible, tube and mold) and calculate from these
numbers the appropriate maximum volume that could be safely removed. Based upon the known
volumes it was determined that the maximum volume that could be removed was 300 cc. A mold
geometry was therefore designed that would allow for indication of the critical withdrawal volume
giving the operator the ability to stop the withdrawal process before jetting of metal from the tube
into the vacuum dome. The mold was of similar style to that shown in section 4.3.2.
4
One consequence of the mold revision was the adjustment of the location of the withdrawal tube
from the center of the mold plate to the side. As a safety precaution it was decided to install a
number other features. First, a wall was installed between the operator and the furnace, plywood
was deemed adequate because its purpose was only to stop small splashes of metal or glass and
fire was not a serious concern. Additionally, a wire screen was constructed to go over the glass
dome to reduce the hazard of implosion sending glass and metal around the area. Finally it was
decided to install a steel sheet inside the dome to deflect the jet of metal, if it should occur, to
prevent direct impingement on the dome.
5.2 Corn Syrup
As the water experiment was an analogue to metal, corn syrup was withdrawn to simulate the
higher viscosity oxide melt. The viscosity of corn syrup is in the range of 2 to 5 Pa·s which gave a
reasonable comparison to the anticipated viscosity for the oxide melt, (between 0.63 and 6.3
Pa·s). The setup for this experiment was identical to that described for water with slight
modification of the tube and mold. In this case rather than using a glued on attachment, the end
of the silica tube was ground to provide an outlet for overflow into the sand mold. This experiment
did not return any unexpected results. The volume removed was kept below the critical amount
and no show stopping issues were identified. A slower fluid velocity was obtained because of the
higher viscosity, but this was expected and by adjusting the needle valve to increase pressure
differential a similar volume flow rate was obtained.
5.3 Aluminum
Five trials were run attempting to withdrawal aluminum from the crucible. The setup and
procedure for each experiment was based upon the result of the previous trials. After five
attempts, the system was successfully validated for aluminum withdrawal.
5.3.1 Setup and Procedure
Setup
The setup for aluminum withdrawal varied slightly between each of the experiments. Table 10
summarizes the data for the trials
5
Table 10: Aluminum Experimental Setup
Trial 1
Trial 2
Trial 3
Trial 4
Trial 5
Alloy Withdrawn
Unknown
Al 319
Al 319
Al 319
Al 319
Metal Weight (g)
Crucible
Volume (cc)
Crucible Type
1120
630
630
1500
1500
475
310
310
800
800
Alumina
Clay-Graphite
Clay-Graphite
Clay- Graphite
Clay- Graphite
Tube
Dimensions (m
-2
x10 )
Tube Material
2.54 OD
x 1.91 ID
x 55.88
Silica
2.54 OD x
1.91 ID x
55.88
Silica
2.54 OD x
2.26 ID x
55.88
Steel
2.54 OD x
2.26 ID x
55.88
Steel
2.54 OD x
2.26 ID x
55.88
Steel
It was initially decided to use a cylindrical alumina crucible with a secondary cylindrical crucible
for backup in the event of a primary crucible failure. However it was decided that minimal risk of
failure containing aluminum and the low amount of damage likely to furnace hardware in the
event of a leak was not worth the cost of the high purity alumina. For this reason clay graphite
crucibles were used for the remainder of the aluminum experiments.
For the trials in which a steel tube was used the bottom 1.52 x10-1 m (6 in) of the tube was coated
with a boron nitride die coating „Lubricoat‟ manufactured by ZYP Coatings for the purpose of
protecting the steel against interaction with the molten aluminum. Liquidus and solidus
temperature for the aluminum alloy used are 610 and 482 °C respectively.
For Trials 3, 4, and 5 the steel tube was insulated with alumina fiber wrapped around the tube to
prevent heat loss during preheat of the tube in the furnace before withdrawal. In Trial 5 the tube
was preheated using a nichrome resistance wire before dropping the tube into the melt.
Alumina
Wool
Insulation
Resistance
Leads
Withdrawal
Tube
Figure 17: Insulation covers resistance heating wire in Trial 5.
6
For each experiment the withdrawal tube was held in place by gluing it into the steel flange using
alumina paste glue. This provided a bond which, though not very strong, allowed a sufficient seal
between the tube and flange to prevent leaks during vacuum application. The top of the tube was
inserted into a shaped cavity in the mold and glued to the mold, again with alumina paste.
Figure 18: Withdrawal tube was glued into sodium silicate bonded sand with alumina paste. Alumina
fiber was used to insulate the tube.
For these trials the seal between the glass vacuum dome and the steel mold plate was made
using a foamed silicone rubber gasket. The mold plate did not get hot enough during withdrawal
to cause concern for burning the gasket. As with the water withdrawal clamping was provided
simply with atmospheric pressure when vacuum was pulled inside the dome.
Withdrawal Procedure
In each of the withdrawal experiments the withdrawal tube was lowered into the furnace using the
linear actuator and allowed to sit inside the furnace with the bottom of the tube just above the
melt surface to preheat before being lowered into the melt. Except for the first experiment, the
outside temperature of the tube was measured using a type K thermocouple at a location 7.62
x10-2 m (3 in) below the steel tube mounting flange. Before dropping the withdrawal tube into the
furnace the temperature of the melt was measured with a type K thermocouple.
7
Table 11: Aluminum Experimental Parameters
Preheat Time (min)
Trial 1
Trial 2
Trial 3
Trial 4
Trial 5
4 + 1 min
submerged
Unknown
96
44
25
21
Preheat Temperature
150
170
112
375-400
(C)
Measurement Location
XXX
33.0
33.0
33.0
33.0
(m x 10-2 Above Tube
Bottom)
Melt Temperature (C)
797
800**
850**
780**
730**
**These temperatures are approximated, the long preheat time prevented measurement directly before
withdrawal.
Just prior to withdrawal the vacuum pump was turned on and the buffer tank was evacuated to
allow a fast drop of pressure in the vacuum dome over the mold. After evacuation of the buffer
volume, the preheated tube was dropped below the surface of the melt to a position between 2.54
x10-3 and 6.45 x10-3 m (0.1 and 0.25 in) above the bottom of the crucible. As with the water
experiments, control of the pressure in the dome was achieved by manual manipulation of a
needle valve on vacuum lines between the buffer tank and the vacuum dome over the mold.
Withdrawal stopped either when the aluminum froze in the withdrawal tube or when the metal
was removed from the crucible to the maximum extent possible.
Immediately following the withdrawal, the tube was removed from the furnace with the linear
actuator. Both insertion and removal took place at a rate of approximately 1.02 x10-2 m (0.4 in)
per second.
5.3.2 Results and Discussion
The results of the aluminum trials are tabulated below. For each of the trials the maximum
expected withdrawal volume was dictated by the mold volume to be 282 cc or about 760g of
aluminum alloy. For Trials 2 and 3, in which the 310 cc clay graphite crucible was used, the
maximum safe withdrawal volume without reaching the critical metal jetting fluid level was only
100 cc or 270g or aluminum. In the table the partial withdrawal distance is the distance above the
initial melt height to which fluid was withdrawn before freezing in the tube.
Table 12: Aluminum Experimental Results Summary
Trial 1
No
Trial 2
No
Trial 3
Yes
Trial 4
No
Trial 5
Yes
XXX
XXX
≈400
XXX
908
Partial Withdrawal
Distance (m x 10-2)
3
41
XXX
31
XXX
Control of Withdrawal
Yes
Yes
No
Yes
Yes
Metal Successfully
Removed
Amount Removed (g)
8
In the first two trials the metal froze in the silica withdrawal tube before reaching the exit of the
tube. The more successful Trial 2 results were achieved by increasing the preheat time and the
rate at which vacuum was applied. In the first experiment the vacuum was applied slowly allowing
a gradual rise of metal in the withdrawal tube. This approach was unsuccessful because it
allowed flow in the tube to be close to plug flow, thereby allowing the top portion of the metal
column to stay in contact with the cold wall of the withdrawal tube, cooling it significantly as it
traveled up until it solidified. Faster application of a pressure differential increased the rate of flow
giving a more parabolic profile to the fluid in the tube. This allowed the center of the metal column
to stay hotter, keeping a path open between the molten reservoir and the surface of the rising
metal. Therefore it was concluded that faster application of vacuum would, at the cost of less
control of the process, help prevent freezing in the tube. For these experiments an electronic
gauge was not yet installed in the system preventing data acquisition, so a quantitative pressure
profile was not measured.
Figure 19: Frozen aluminum in silica withdrawal tube showing only one inch of removal past the
initial melt height.
9
Though outcome of the second experiment was better than that of the first, the result was still a
frozen plug in the tube rather than a successful withdrawal. After two failures with a silica tube,
the change to steel was made in hopes that the higher thermal conductivity of steel would give a
higher preheat temperature on the portion of the tube that was not inserted in the furnace during
the preheat. Also, though there was limited exposure time, the silica was reduced by aluminum
metal encouraging a switch of tube materials.
The change to steel had the desired effect. Using a similar rate of vacuum application as in Trial
2, aluminum was withdrawn completely from the crucible and no metal was frozen in the
withdrawal tube in Trial 3. However, as mentioned, the faster vacuum rate prevented shutting the
flow off fast enough and a metal jet was experienced similar to what was seen with the water
withdrawal. Figure 20 depicts the results of the jet which shattered the glass dome from thermal
shock. Use of a protective wire screen prevented any glass or metal from escaping after fracture
and implosion.
Figure 20: Glass dome after cracking due to thermal shock from molten aluminum splash.
After replacement of the glass dome it was decided to use a larger volume of metal to prevent
another jetting incident. Using a larger crucible would allow withdrawal of more material without
draining the crucible and reaching the critical fluid height. Trial 4 again used a steel tube but
10
freezing occurred in the tube in the section underneath the mounting flange. This area was
expected to remain the coolest section because of the larger thermal mass around that section of
the tube.
The setup for Trial 5 was modified as shown in Figure 17 to prevent freezing under the flange.
Because the tube was preheated more uniformly, and to a high temperature, this experiment had
perfect results. The mold was completely filled with metal at an average flow rate of 69.3 cc. per
second and after the pressure differential was removed the column in the withdrawal tube
dropped back into the crucible leaving a completely clean withdrawal tube. Some aluminum oxide
dross stuck to the bottom of the tube after removal from the melt; however, the boron nitride
coating prevented strong adhesion, allowing it to be removed in one solid piece by peeling with
minimal force.
5.3.3 Conclusions and Modifications
Based on the results obtained in the last aluminum experiment it was shown that aluminum could
successfully be withdrawn using the system. Additionally, the hardware was not damaged
allowing use for multiple trials if desired. However, because aluminum handling was not an
objective of the project, it was decided to continue to a higher temperature melt without spending
the effort to optimize the device for aluminum withdrawal.
One significant issue with the aluminum experiments was the jetting of metal and risk to the
operator and system hardware. Two possible reasons for the metal jet are proposed. First the
result could be due to a similar process as proposed in section 5.1.2. Alternately, the issue could
be because of the formation of a metal skin. In the metal skin case, a thin skin of frozen metal
forms across the tube cross section. Because the metal is still very hot and not very strong, it
tears upon application of sufficient pressure differential allowing metal to shoot up the tube.
Conceivably both situations could occur where a skin forms and tears first and after the
uncontrolled flow of metal up the tube the level of metal in the crucible is too low allowing the
second type of jetting event.
Regardless of the jetting cause, it was determined that due to the difficulty of controlling the metal
flow and the lack of quantitative data regarding the critical measurement of rate of pressure drop
in the vacuum chamber it was necessary to create a better container than glass for holding
vacuum and to install a means of measuring pressure above the mold. A metallic vacuum
container with ports for installing measurement devices was therefore designed and ordered.
11
5.4 Ferrosilicon
After proving the system with aluminum it was decided to attempt withdrawal with a ferrosilicon
melt similar in composition to the expected product of the MRE process. Two trials were
attempted with a 75 wt% silicon 25% iron alloy.
5.4.1 Setup and Procedure
Setup
It was determined during testing of the furnace that the setup of the furnace tube used for
aluminum was not acceptable for higher temperature experiments. In the aluminum trials the
furnace tube passed completely through the borehole of the furnace and was capped with an
alumina fiber board. The exposed sides of the tube were covered in alumina wool insulation. This
setup did not allow the furnace to reach the requisite 1600 °C for further testing so a modification
was made. A new alumina plug was made and the furnace tube was shortened to be completely
enclosed in the furnace rather than pass though the furnace roof. This allowed the furnace to
easily reach 1600 °C.
The parameters used for the withdrawal experiments are listed in Table 13.
Table 13: Ferrosilicon Experimental Setup
Trial 1
Trial 2
Alloy Withdrawn
75 Si - 25 Fe
75 Si - 25 Fe
Metal Weight (g)
Crucible Volume (cc)
1330
500
1252
500
Crucible Type
Tube Dimensions
(m x10-2)
Alumina
2.21 OD x 1.91
ID x 55.88
Alumina
2.21 OD x 1.91
ID x 55.34
Tube Material
Fused Silica
Fused Silica
For ferrosilicon withdrawal it was decided use an alumina crucible because of its higher
temperature capability relative to the clay graphite used with aluminum. The dual containment
strategy was used, nesting a 500 cc crucible inside a 750 cc crucible with alumina sand between
crucibles. Fused silica withdrawal tubes were used because of their thermal shock resistance.
Though silica is not a compatible material with molten regolith it was expected to provide an
effective material for validation of ferrosilicon withdrawal.
As with the aluminum experiments the tube was held into and sealed against the steel flange with
alumina paste glue. Because the outside diameter of the withdrawal tubes was smaller for these
experiments than the aluminum experiments a short steel sleeve was used as a bushing to
provide alignment of the tube in the flange which was made to fit a one inch diameter tube. The
12
tube was inserted and glued into the sleeve which was then glued into the flange. A glass dome
was used as vacuum chamber because the aluminum dome designed was not yet finished being
machined. Again, a foamed rubber gasket was the sealing method. At the top of the withdrawal
tube a 2.54 x10-2 m (1 in) thick piece of ceramic fiber wool was used to insulate the withdrawal
tube and protect the glass dome from radiant heat from the tube (similar to Figure 18).
Melting of the charge was done by filling the crucible with granular metallurgical ferrosilicon,
melting, and making additions to the melt to bring the liquid pool to the full mass specified.
Granular material was simply dropped through the bore hole in the alumina plug and allowed to
fall into the crucible. A funnel was used to aid pouring of the ferrosilicon into the crucible. A flow of
argon was kept over top of the ferrosilicon to help prevent oxidation of the material. This was a
concern because of the large surface area to volume ratio of the granular material; a 9.5 x10-3 m
(0.375 in) mesh product was used. After the last addition to the melt was made the melt was held
for 30 and 45 minutes respectively for Trials 1 and 2 to allow the temperature to stabilize.
Withdrawal Procedure
Table 14: Ferrosilicon Experimental Parameters
Trial 1
Trial 2
Preheat Time (min)
0
0
Melt Temperature (C)**
1460
1460
** The melt temperature was approximate; it was not directly measured because of the concern for damage
to the type B thermocouple required for temperature measurements in this range. It was later discovered
that the calibration of the furnace control TC was off indicating a withdrawal considerably below the
expected 1600 °C withdrawal temperature. It is estimated that the temperature listed is accurate +- 10 °C.
The procedure used for withdrawal of ferrosilicon was simpler than that used for aluminum
because the preheat step was eliminated. After reaching full vacuum on the buffer tank the tube
was dropped into the melt, approximately 7 x10-3 m (0.30 in) above the crucible bottom and
vacuum was immediately applied by manually opening the needle valve. The silica tubes used
allowed observation of the fluid level in the tube as well as the outlet for control of the vacuum
rate. Unfortunately, the metallic vacuum dome was not complete when these trials were run so a
time-pressure curve was not measured.
5.4.2 Results and Discussion
Both of the experiments successfully removed ferrosilicon from the crucible.
Table 15: Ferrosilicon Experimental Results Summary
Metal Successfully
Removed
Amount Removed (g)
Control of Withdrawal
13
Trial 1
Yes
Trial 2
Yes
≈155
Yes
571
No
Because there was no preheat for either trial the main difference between the two was the rate of
vacuum application. In the second trial the vacuum was applied considerably faster than in the
first test. The change was made because of the significant amount of freezing seen in the first
trial; the tube was actually completely full resulting in a frozen plug being removed from the
furnace. This meant that not much ferrosilicon was removed to the mold and that the process
would not be repeatable with the same hardware because of the fully closed withdrawal tube.
In the second test vacuum was applied much faster. Though a quantitative measurement of
pressure is not available, video recorded of the withdrawal showed that prior to exit from the
withdrawal tube the fluid front was moving at 0.25 m/s (10 in/s) in the second trial as compared
-2
to 3.81 x10 m/s (1.5 in/s) in the first trial. This corresponds to flowrates of approximately 10.9
and 72.4 cc per second. The faster withdrawal rate resulted in significantly less freezing in the
tube for the second experiment. The tube was mostly empty after withdrawal; a thin skin of
ferrosilicon froze on the inside wall approximately 3 x10-3 m (0.12 in) thick at the thickest point.
This point was underneath the steel flange, which would be expected as this area had the largest
thermal mass and there was little insulation between the tube and the flange to prevent
conduction of heat between the two.
Because of the fast rate of withdrawal in the second trial the withdrawal was uncontrolled
resulting in a jet of metal at the end of withdrawal. As with the aluminum incident the metal
splashed around the glass vacuum chamber resulting in a fracture which destroyed the container.
Figure 21 shows the results of the withdrawal which despite the splashing were considered
successful as they proved that ferrosilicon could be removed from the crucible without complete
freezing.
14
Figure 21: Ferrosilicon immediately after withdrawal in the second experiment.
One significant complication with withdrawal was the oxidation of the melt. Though the ferrosilicon
was melted under an argon flow the furnace tube was not completely sealed allowing oxygen
leakage during melting and the period in which additions were made. Because of heavy oxidation
some pieces of ferrosilicon did not completely fuse resulting in a thick mat of oxides in the
crucible. This probably helped contribute to the splashing in the second experiment since the
oxide layer was porous and probably allowed air to reach the tube inlet with a lower withdrawal
volume than expected.
Another result that required addressing in future experiments was the fact that in both trials the
fused silica withdrawal tube broke from thermal shock after removal of the metal. After withdrawal
the tube was allowed to stay lowered in the furnace to encourage gravity driven draining. The
middle portion of the tube was not immersed in the furnace or contained in the flange but was
exposed to open air during and after withdrawal. This portion of the tube was not insulated and
cooled both by convection and radiation. One minute 40 seconds, and 30 seconds respectively
after the end of each withdrawal this area shattered from thermal stresses during cooling.
15
Because of the large amount of oxidation present in the crucible in the first two experiments a
third attempt at withdrawing ferrosilicon was made. This trial used a larger melt contained in a
clay graphite crucible. During heating the crucible cracked resulting in significant damage to the
furnace causing an abort to the experiment.
5.4.3 Conclusions and Modifications
These two experiments proved that ferrosilicon could successfully be withdrawn from a crucible
using the pressure differential method. Given its use in industrial practice it was expected that this
method would be viable but these tests proved the method using the given configuration.
The most significant result of this experiment was the proof that once again rate of vacuum
application was the critical parameter to control the success of withdrawal. Even without any
preheating, the withdrawal of Trial 2 was successful at 1460 °C, considerably below the expected
operational temperature of an MRE cell. Heating further would only improve fluidity and reduce
freezing concerns.
The fracture of the silica withdrawal tubes suggests that because fused silica has high resistance
to thermal shock for an oxide ceramic that this class of material is not a viable choice for the
withdrawal tube. Insulation of the tube is an option but if the tube is desired to be preheated in the
furnace by radiant heating this insulation would slow the heating and therefore the entire
withdrawal cycle reducing the functional time of the proposed electrolysis cell. Though a
resistance heated configuration could be designed to both allow preheating and prevent thermal
shock to allow oxide ceramic use this complication was avoided as it was not the primary
objective of the project. The design of such a setup is left to future studies using this method.
Given the relative success of the second ferrosilicon withdrawal and demonstrated risk to
equipment inherent with experimentation it was decided to direct work at removing a glass from
the cell before addressing other concerns of the ferrosilicon withdrawal process. Concerns about
the viscosity of a silicate melt, and its dramatic increase with lowered temperature, motivated
modifications to the system to further reduce the possibility of freezing in the tube. First an
extension was made to allow the mold plate to drop closer to the furnace; this distance was
previously limited by the actuator travel. Addition of the extension removed four inches from the
minimum tube length. Also the vacuum system was modified, doubling the volume of the buffer
tank and changing the valve from a needle valve to a ball valve. It was decided that while a
needle valve gave more control it opened too slowly, making fast withdrawal difficult with little
benefit to the extra control.
16
5.5 Low Viscosity Glass
Before attempting withdrawal of a depleted regolith melt it was decided to withdrawal a glassy
melt with similar viscosity profile to the depleted regolith at a lower range of temperatures. This
would enable investigating the viability of the pressure differential method for glass at a lower
temperature than required for MRE. Using Sciglass software, a composition similar to a depleted
regolith melt glass was identified by project team members which had a similar viscosity but at
200 °C lower temperature, meaning that the viscosity of depleted regolith at 1600 °C was the
same as the modified melt at 1400 °C22. The general relationship between temperature and
viscosity is the same for these melts so the viscosities are equivalent at roughly 200 degrees in
temperature difference for the entire range relevant to withdrawal.
Table 16: Depleted Regolith vs. Modified Low Viscosity Melt Composition
Depleted
Regolith
Low
Viscosity
Melt
Component
Wt%
Wt%
SiO2
37.8
38
Al2O3
28.5
13
MgO
15.4
11
CaO
18.3
38
Total
100
100
The benefits of reduced temperature were lower risk of damage to the furnace, and the ability to
use a low cost withdrawal tube, namely steel rather than a ceramic or refractory metal.
5.5.1 Setup and Procedure
Setup
Table 17: Low Viscosity Glass Experimental Setup
Trial 1
Glass Charge Weight (g)
Crucible Volume (cc)
1134
500
Crucible Type
Tube Dimensions
(m x 10-2)
Magnesia
2.21 OD x 1.91
ID x 45.72
Tube Material
4130 Steel
For this experiment the crucible was changed to magnesia. The higher calcium oxide and lower
silica content of the melt increased the basicity of the melt so a basic oxide was chosen for
containment to prevent attack by the melt. The crucible for this experiment was 96% MgO + 3%
Yt2O3 supplied by Aremco. It was contained within a second magnesia crucible also from Aremco.
17
Steel was sufficient for use as the withdrawal tube material at 1400 °C. It was at the limit of its
temperature capability but under its melting point. The length was decreased from 0.55 to 0.46 m
(22 to 18 in) because of the extension plate as described in the previous section. The entire
inside and outside of the tube was coated with Lubricoat boron nitride with a thickness of
-4
approximately 1.27 x10 m (0.005 in) thick to help reduce interaction with the melt and prevent
wetting of the tube by the glass. For this experiment the flange was slightly modified. It was bored
out to allow insertion of a ceramic fiber bushing into which the steel tube was glued. This allowed
a lower conductivity from the tube to the flange to prevent heat loss at this point during
withdrawal.
The shorter length, and therefore smaller volume, of the withdrawal tube allowed more of the melt
to be removed before reaching the critical splashing level so a new mold similar to the previous
mold was designed with a larger volume, 283 cc vs. 212 cc.
For this experiment the newly fabricated aluminum vacuum dome was used as a replacement to
the second broken glass dome. This new container had two Vycor glass windows to allow the
operator to view the process while filming the withdrawal through the second port. The dome also
included two ports into which were inserted a pressure transducer (Measurement Specialties
US300 series) and a thermocouple for measuring temperature of the tube inside the dome during
withdrawal. A type K thermocouple was placed between the tube and the mold at a position 0.05
m (2 in) below the top of the tube. A fine gauge (1.27 x10-4 m, 0.005 in diameter) type S
thermocouple was attached to the withdrawal tube at a position 0.20 m (8 in) above the bottom of
the tube using alumina paste glue. These instruments were wired to two data acquisition boards
to allow monitoring of pressure and temperature during the withdrawal process. The inside of the
dome was protected against a jet of hot fluid by a ceramic fiber liner.
Withdrawal Procedure
To prepare the melt the appropriate ceramic powders (Sigma Aldrich) were mixed in plastic
containers. The crucible was loaded with 470g of gently compacted ceramic powder. The
remaining weight was pressed by hand into alumina crucibles and sintered at 900 °C for 1 hr. The
resulting lightly bonded mass was broken into pieces and fed into the crucible after melting of the
initial charge to 1400 °C. Charges of 302, 150 and 212 g of material were made attempting to add
the maximum amount to the crucible without overflowing; the difficulty of observation inside the
furnace prevented accurate charging. Additions were made by simply using a steel funnel to drop
oxide pieces through the furnace plug into the crucible and pressing the charged mass into the
melt with an alumina rod. After addition of the full weight of oxide, the melt was allowed to
homogenize for 1 hour. Just prior to withdrawal a type B thermocouple was immersed in the melt
18
and quickly reached an output indicating between 1380 and 1390 °C. The actual temperature
may have been slightly higher but the thermocouple was not left in the melt to obtain a long term
reading because of concern for damage to the alumina sheath and destruction of the
thermocouple itself.
Table 18: Low Viscosity Glass Experimental Parameters
Trial 1
Preheat Time (min)
1
Preheat Temperature (C)
Unmeasured
-2
Measurement Location (m x 10
Above Tube Bottom)
Melt Temperature (C)
20.3 (8 in)
1380-1390
The tube was preheated at a position approximately 2.54 x 10-2 m (1 in) above the melt surface
-2
for 1 minute before being dropped to 6.35 x 10 m (0.25 in) above the bottom of the crucible.
Immediately after reaching the withdrawal position the ball valve connecting the aluminum
vacuum dome to the previously evacuated tank was opened allowing the pressure in the dome to
be reduced and molten oxide to be withdrawn. Following withdrawal the tube was driven at 2.54
x10 -2 m/s (1 in/s) back to the preheat height and held for 2 minutes to allow the molten glass in
the tube to drain back into the crucible. After draining, the tube was withdrawn 0.7 m (24 in) to the
starting position, again at 2.54 x10 -2 m/s (1 in/s)..
5.5.2 Results and Discussion
Table 19: Low Viscosity Glass Experimental Results Summary
Glass Successfully
Removed
Trial 1
Yes
Amount Removed (g)
529
Control of Withdrawal
No
The general result of the experiment was that the oxide glass was successfully removed from the
crucible using the pressure differential device. The withdrawal was run until all available oxide
was removed, resulting in uncontrolled splashing of the melt out of the withdrawal tube as was
seen in previous runs with metal. Due to the higher viscosity of the oxide melt the splashing was
not as significant a problem. The ejected material simply formed small blobs which stuck to the
inside of the ceramic fiber liner constructed specifically for the purpose of protecting against
splashing.
During cooling both magnesia crucibles cracked, but none of the oxide leaked out indicating that
cracking occurred after freezing of the oxide.
19
Based on video footage of the experiment, the withdrawal of molten material contained two
stages. During the first stage the material withdrawn was viscous and oozed out of the withdrawal
tube. During the second stage the material flowed easily out of the tube and into the mold. The
initial material removed piled up at the exit of the withdrawal tube while the easily flowing oxide
ran into the mold.
It is proposed that the initial material was more viscous simply because it cooled during its
passage up the withdrawal tube; while during the later stage of withdrawal the tube was already
hot allowing the oxide to stay hot and flow easily.
Figure 22: Results of low viscosity glass withdrawal. Mass around exit of tube had a high viscosity
during withdrawal while the glass in the mold flowed easily. This difference is probably due to
chilling of the initial material by the tube.
A plot of pressure and temperature vs. time is given in Figure 23 which gives an indication of the
operating parameters required for oxide withdrawal. Unfortunately the type S thermocouple
became disconnected during the experiment preventing any data from being recorded regarding
the temperature of the lower portion of the tube during preheating.
20
16
700
14
600
12
500
10
400
8
300
6
200
Pressure (PSI)
Temp (C)
800
4
Temperature
Pressure
100
2
0
0
100
120
140
160
180
Time (s)
200
220
240
Figure 23: Pressure and temperature data from low viscosity glass withdrawal. Pressure
measurement is absolute pressure in vacuum chamber; temperature was recorded at two inches
below the top of the tube in between the sand mold and the tube. Synchronization of the data is
approximate because two different boards were used for signal acquisition. Time boundary is
arbitrary; zero is the start of data acquisition.
An important observation of the process was that the tube was considerably closed by the
freezing of an oxide layer on the inside of the tube. The tube was sectioned for investigation. At
the narrowest point the tube was constricted to an open diameter of only 4 x10-3 m (0.16 in); of
note this point was directly under the steel flange, so this result was not surprising.
Several other features of interest were noted after cooling of the melt, namely the metallic coating
on the top of the withdrawn part, the gas porosity in the top of the glass and a pattern of three
different colors observed both in the residual oxide in the crucible and in the frozen plug in the
tube. Efforts to identify differences in composition using EDS did not result in conclusive evidence
of any disparity between the color of the solidified glass. Nor did a brief investigation positively
identify the thin coating on the glass surface. It is postulated that the metallic coating is a reaction
product with a reducing agent which also produces the gas bubbles. The only likely reducing
agent is boron nitride from the coating; however thermodynamics suggests no reactions with any
of the components of the melt.
It was also noted that there was a significant lack of homogeneity in the melt with regards to
magnesia content with individual particles of magnesia being easily identified under both the
optical and electron microscope. Improved melting procedure would prevent this issue.
21
5.5.3 Conclusions and Modifications
This trial proved that withdrawal of an oxide melt is possible using the pressure differential
method. The similar viscosity profile of the melt used to that of depleted regolith suggested that
an experiment with the actual target composition would be successful as well. Concerns brought
up during this trial were the high viscosity of the initial material through the tube and the
associated plug of glass left after withdrawal. Addressing this concern could be done by simply
preheating the tube more before withdrawal; however, because of the loss of the lower
thermocouple data quantitative comparison of preheat temperatures was not possible. It was
determined therefore to simply increase the preheat time; as a longer time at a higher furnace
temperature would greatly increase preheating of the tube.
The other concern brought up was the inhomogenity of the melt. The poor mixing was a result of
both a large starting size of the magnesia particles used in the oxide mixture, lack of agitation of
the melt and a short homogenization time. Future experiments would alleviate this problem by
grinding the MgO before melting, stirring and for the depleted regolith experiments using a higher
melting temperature. From a validity standpoint it was not concerning that the magnesium oxide
was not fully dissolved into the melt. In fact, the melt was probably more viscous than expected
as a result of an effectively higher silica content and potentially containing unmelted magnesia
particles. Better mixing would probably have resulted in an easier withdrawal.
Because the low viscosity glass was not the target of the investigation it was decided that after
only one validation experiment the best course of action was to proceed to the depleted regolith
composition.
5.6 Depleted Regolith
During molten regolith electrolysis, as electrolysis proceeds the melting point of the molten oxide
electrolyte increases. As electrolysis and withdrawal are set for an operating point of 1600 °C, the
electrolyte must have a melting point below this temperature. Based upon the JSC-1A
composition this point corresponds to a composition in which iron oxide is entirely reduced from
the melt and roughly half of the silicon is reduced. The composition for depleted regolith
withdrawal was therefore chosen to be that given in Table 2.
5.6.1 Setup and Procedure
Setup
22
Table 20: Depleted Regolith Experimental Setup
Trial 1
Glass Charge Weight (g)
Crucible Volume (ml)
1649
500
Crucible Type
Tube Dimensions
(m x 10 -2)
Alumina
2.21 OD x 1.91
ID x 45.72
Tube Material
99.9% Mo
The setup for depleted regolith withdrawal was identical to that used for the low viscosity glass
withdrawal with the modification of a few factors. The crucibles were changed back to alumina,
reflecting the switch back to a higher silica melt. Additionally, because of the higher withdrawal
temperature the withdrawal tube used was molybdenum rather than steel; again the tube was
fully coated in Boron Nitride for reducing wetting by the melt. The average coating thickness was
2.0 x10 -4 m (0.008 in).
The same instrumentation was used as in the low viscosity glass withdrawal, with the addition of
a second type S thermocouple which was placed in the mold to measure the melt temperature at
the outlet of the tube.
Withdrawal Procedure
As with the low viscosity glass, the melt was prepared by first mixing powders in plastic
containers. The primary crucible was filled with 735g of powder and loaded into the furnace. The
remaining powder was hand pressed into an alumina container and sintered at 1000 °C for two
hours. After melting the initial charge to 1600 °C additions were made in weights of 350, 175 and
389g each time pressing the newly added pieces into the melt with an alumina rod. After the last
addition the melt was allowed to homogenize for roughly four hours before withdrawal. During the
holding period the melt was stirred several times with an alumina rod. Immediately prior to
withdrawal the melt temperature was measured with a type B thermocouple to be 1610 °C.
Table 21: Depleted Regolith Experimental Parameters
Trial 1
Preheat Time (min)
2
360, at 8” TC
Preheat Temperature (C)
-2
Measurement Location (m x 10
Above Tube Bottom)
Melt Temperature (C)
20.3 (8 in)
1610
Preheating of the tube was done for two minutes with the bottom of the tube 6.4 x10-3 m (0.25 in)
above the melt surface. After withdrawing the melt the tube was driven to the preheat position to
drain for two minutes. All actuator motion was driven at one inch per second average velocity.
23
5.6.2 Results and Discussion
Table 22: Depleted Regolith Experimental Results Summary
Trial 1
Glass Successfully
Removed
Amount Removed (g)
Yes
Control of Withdrawal
Yes
767
The withdrawal of depleted regolith was successful; the complete amount of fluid in the crucible
was removed to the sand mold. Again a pressure profile was recorded during the experiment.
Unfortunately while the thermocouples did not become disconnected during the withdrawal the
recorded output was not clear because of significant noise in the system. Also, the type S
thermocouple on the tube was dislodged by scraping the furnace plug during insertion into the
furnace so the measurement recorded after the significant fluctuations occur is not accurate for
the actual tube temperature.
800
Tube Low
700
Tube High
600
Temp (C)
500
400
300
200
100
0
0
50
100
Time (s)
150
200
250
Figure 24: Temperature data recorded from tube thermocouples, melt thermocouple did not yield
any significant reading. The fluctuations in thermocouple output began around time = 175 seconds,
right as the lower tube thermocouple was broken, it is likely that this caused interference with the
data acquisition board. Time is zeroed to first downward motion of the actuator. Withdrawal occurs
at time = 170 seconds.
24
16
14
Pressure (PSI)
12
10
8
6
4
2
Depleted Regolith
Low Viscosity Glass
0
165
170
175
180
Time (s)
185
190
195
Figure 25: Pressure profile from depleted regolith experiment compared to that of the low viscosity
glass. The low viscosity glass data has been translated to overlay.
Graphical representation of the pressure during withdrawal shows that the process was much
easier for depleted regolith than for the low viscosity glass. This may have been due to higher
preheating or a lower actual viscosity of the melt. Both the pressure differential and the time
required for full withdrawal were much less in this experiment. Because of the exponential
relationship between viscosity and temperature, greater preheating of the tube was likely a strong
factor in influencing this result. The average flow rate for the experiment was 37cc per second.
25
Figure 26: Depleted regolith withdrawal results.
Pictorial evidence of the withdrawal is shown in Figure 26. The glass ingot cast fractured from
thermal stress during cooling in the mold after withdrawal. For the present investigation this is not
a concern. The conclusion of withdrawal was not adequately indicated by the volumetric method
intended by the mold design. Jetting occurred as too much liquid was removed, however this was
an operator error rather than a failure of the system so more caution and a better indicator would
otherwise have given positive results.
The boron nitride coating flaked off of the withdrawal tube allowing it to undergo oxidation during
the trial. Volatile molybdenum oxide deposited on the underside of the withdrawal plate. The
depth of oxidation was insignificant given the minimal exposure of the tube; measurement did not
indicate any change in dimension. The inside of the tube was coated with a layer of depleted
regolith glass. As with the previous experiment the thickest part of the glass was under the steel
flange. The glass coating thickness was difficult to determine since the tube was not sectioned,
-3
but at the narrowest point the tube diameter was approximately 9.5 x10 m (0.375 in). The total
volume of glass frozen in the tube was 37 cc of a total 130 cc tube volume. Most of this mass was
26
concentrated at the upper portion of the tube; the section that remained in the furnace while
draining was coated with only a film of glass. There was no noticeable attack on the outside of the
tube by the glass.
5.6.3 Conclusions and Modifications
This experiment proved that depleted regolith could be withdrawn from a reactor cell using the
pressure differential method. Molten oxide was withdrawn completely without blockage of the
withdrawal tube. The fluidity of the depleted regolith melt was confirmed at the 1600 °C process
temperature used for withdrawal. It was shown that the molybdenum withdrawal tube stood up to
brief temperature exposure in an oxidizing environment leaving the tube usable for further trials.
The time-pressure profile required for the withdrawal was obtained. The preheat temperature of
the tube was measured; unfortunately no temperature data was recorded during the withdrawal.
The reason for thermocouple failure in the experiment is likely because of electronic instability of
the data acquisition board that occurred upon failure of the lower tube thermocouple. When
mechanical failure occurred, the thermocouple shorted to the tube. This was shown to influence
the reading of other thermocouples in later investigations of the cause.
Control of the withdrawal process and elimination of jetting could be achieved by using
computerized control of the pressure in the vacuum dome rather than manual control to achieve a
specified time-pressure curve. It was not the objective of this work to develop such an automated
system, but rather define generally the parameters required; therefore it was decided not to
engage in efforts to better control vacuum chamber pressure for better performance.
Given the success of this experiment, it was decided to continue to the final challenge of dual
withdrawal of both a ferrosilicon melt and an oxide melt. Withdrawal of a two layered melt was the
ultimate objective of this experiment as it represents the type of configuration expected in a
molten regolith electrolysis cell.
5.7 Dual Withdrawal of Molten Regolith and Ferrosilicon
Dual withdrawal implies the simultaneous removal of both a reduced metallic layer and a depleted
oxide electrolyte layer from the MRE cell. While it is possible to envision separate ways to remove
the metal and oxide layers, it would be greatly desirable from a simplicity and weight savings
standpoint to use the same mechanism for removal of both the metallic and oxide products of the
cell. The proposed cell configuration would require removal of the metal layer from underneath
the layer of oxide. While two different tubes could be used to separately handle each fluid layer it
would be desirable to use a single tube for both components. An additional benefit of running dual
withdrawal with a single tube would be to allow preheating of the withdrawal tube by removal of
27
ferrosilicon before withdrawal of molten oxide. Given the lower viscosity and melting point of
ferrosilicon this would not create concerns for freezing and would allow a large amount of preheat
to prevent the oxide from freezing. Since both ferrosilicon and depleted regolith were shown to be
removable with the system, the largest concern with the dual withdrawal process is the difficult
materials compatibility requirements imposed on the tube. Any tube would have to be able to
withstand contact with both components and be resistant to thermal shock. The work on this area
was therefore directed at evaluating several materials choices to determine the best option for
use in future work on developing the process.
5.7.1 Setup and Procedure
The same basic process was repeated with two different withdrawal tubes. Materials choices
were molybdenum, extruded silicon carbide (Hexoloy SE, St. Gobain), and hot pressed boron
nitride (HBC, Momentive). The setup for the various experiments is summarized in Table 23.
Setup
Table 23: Dual Withdrawal Experimental Setup
Trial 1
Trial 2
Trial 3
Trial 4
Trial 5
Metal Weight (g)
40-60 wt%
Fe-Si +
Depleted
Regolith
(Table 2)
767
40-60 wt%
Fe-Si +
Depleted
Regolith
(Table 2)
767
40-60 wt%
Fe-Si +
Depleted
Regolith
(Table 2)
767
40-60 wt%
Fe-Si +
Depleted
Regolith
(Table 2)
767
40-60 wt%
Fe-Si +
Depleted
Regolith
(Table 2)
767
Oxide Weight (g)
652
652
652
652
652
Crucible Volume
(cc)
Crucible Type
Tube Dimensions
(m x10-2)
500
500
500
500
500
Alumina
2.21 OD x
1.91 ID x
45.72
Molybdenum
(99.9%)
Alumina
2.20 OD x
1.70 ID x
45.75
SiC (extruded
and sintered)
Alumina
2.20 OD x
1.70 ID x
45.75
SiC (extruded
and sintered)
Alumina
2.54 OD x
1.91 ID x
45.72
BN (high
purity)
Alumina
2.54 OD x
1.91 ID x
45.72
BN (high
purity)
Alloy Withdrawn
Tube Material
For trials 1, 2 and 3 instrumentation was identical to that of the depleted regolith withdrawal
experiment with the replacement of the type K thermocouple with a type S. This thermocouple
was also lowered approximately 0.01 m and placed on the exposed side of the tube rather than in
between the tube and mold. In trials 4 and 5 this thermocouple was switched back to type K in an
attempt to avoid difficulty experienced with data acquisition using a delicate, low output sensor.
An additional type K was used to measure temperature at the lower point on the tube.
28
Withdrawal Procedure
Table 24: Dual Withdrawal Experimental Parameters
Trial 1
Preheat Time (min)
Preheat Temperature (C)
Measurement Location
(m x 10-2 Above Tube
Bottom)
Melt Temperature (C)
3
410
20.3 (8 in)
1590-1600
Trial 2
Trial 3
Trial 4
1
30
5
188
415
330
20.3 (8 in) 20.3 (8 in) 20.3 (8 in)
15901600
16001610
15901600
Trial 5
5
300
20.3 (8 in)
>1570
In each of the experiments the preheat temperature increased only at the lower measurement
point. The temperature at the top of the tube was between 50 and 60 °C because of hot air
coming off of the furnace during melting but did not increase significantly during the preheat.
Failure of the data acquisition system during Trial 3 prevented measurement of the upper tube
temperature after a long preheat.
5.7.2 Results and Discussion
Table 25 summarizes the results from the dual withdrawal experiments.
Table 25: Dual Withdrawal Experimental Results Summary
Trial 1
Yes
Trial 2
No
Trial 3
No
Trial 4
Yes
Trial 5
Yes
Total mass Removed (g)
445
XXX
XXX
1056
465
Control of Withdrawal
No
No
No
Yes
Yes
Material Successfully
Removed
Trial 1 was successful in removing both ferrosilicon and depleted regolith from the crucible;
however the molybdenum tube used dissolved very rapidly in the ferrosilicon melt destroying the
lower 0.08 m of the tube in only 40 seconds of exposure. This effectively raised the critical height
for air bubble aspiration and resulted in a splash inside the vacuum dome before the target
withdrawal volume. The tube was rendered unusable after one experiment
Trials 2 and 3 attempted to use silicon carbide tubes since the material is known to have good
shock resistance and exhibited adequate corrosion resistance for short term exposure in
corrosion trials. The tube in trial two failed from thermal shock during application of vacuum
causing the tube to shatter. Only several small drops of ferrosilicon were transferred to the mold.
Because of a furnace electrical failure right before withdrawal preheating time for this trial was
limited to only 1 minute allowing the tube to reach only 188 °C at the measured point before
withdrawal. Given the result of this trial the preheating time was increased to 30 minutes for the
second SiC test. In this test an unknown failure of the data acquisition hardware prevented
29
accurate measurement of the tube temperature before and during withdrawal; however prior to
withdrawal the tube temperature was measured with another thermocouple at 415 °C. Even with
this larger preheat the tube still failed immediately upon vacuum application. In both cases the
point of failure was directly below the tube mounting flange.
Figure 27: Cross section of withdrawn ingot from Trial 1 showing both ferrosilicon layer and an
oxide layer.
Trials 4 resulted in successful removal of both depleted regolith and ferrosilicon from the crucible.
After the trial the boron nitride tube contained frozen material, constricting the tube diameter to
approximately 9.8 x 10-3 m (0.25 in). The lower portion of the tube which was heated during
draining was mostly empty and only small drops of glass stuck to the outside of the tube. The
tube appeared to be on fire immediately after removal from the furnace, with flames exiting the
bottom of the tube. However as the tube cooled the flames stopped. Temperature data from the
trial is given in Appendix B. The flow rate was approximately 76 cc/s.
Trial 5 was run immediately following Trial 4 and therefore the tube was constricted from that
experiment. The test was cut short by the melting of the vacuum hose used to pull vacuum on the
dome. Prior to melting, a considerable amount of ferrosilicon was withdrawn, heating top of the
tube to 1248 °C, however the material froze in the tube completely blocking the withdrawal tube
from 1.52 x10-1 m (6 in) from the bottom to 7.62 x10 -2 m (3 in) from the top of the tube. The flow
rate for this trial was about 30 cc/s. After two trials the withdrawal tube did not have measurable
dimensional change, though the surface of the tube was covered with a boron oxide coating.
5.7.3 Conclusions and Modifications
Based upon the results of the trials described, refractory metals are confirmed to be unusable for
withdrawing ferrosilicon. Silicon carbide has inadequate resistance to thermal shock for the setup
30
used, even with significant preheating. The material may have value if it can be uniformly heated
to high temperature by another means than used in this experiment.
Hot pressed boron nitride is a viable material for dual withdrawal. It was seen that the material
had no problems with thermal shock, even with a measured temperature rise of 990 °C/s at the
top of the tube during withdrawal. Corrosion of the tube was not visible, though some oxidation
did take place. Because freezing and plugging of the tube was a problem even though the tube
became very hot (over 1200 °C) during the withdrawal, a better means of preheating the tube, or
heating after withdrawal, will be required to prevent blockage for multiple uses.
31
Section 6.
General Discussion and Thoughts on
Lunar Implementation
An overall evaluation of the general experimental results indicates that the pressure differential
withdrawal method is a feasible means of removing molten materials from a high temperature
molten regolith electrolysis cell. It was shown that a dual layered melt could be withdrawn from a
furnace and all understanding indicates that scaling of the process would be possible; perhaps
even making the process easier. There were however several issues noted that made the
process challenging and will require significant effort if the method is to be implemented as a
method in lunar operation. In the author‟s proposed order of importance these issues requiring
direct attention are listed below.



Establishing a preheating/ postheating method for the withdrawal tube
Modification of the system for pressurized removal
Development of an automated pressure differential control system
6.1 Focus Areas for Further Development
6.1.1 Preheating/ Postheating Method
One of the critical factors determining the success of withdrawal was the temperature of the tube
when vacuum was applied to remove the melt. For both aluminum and the glass trials it was
noted that having a larger preheat temperature was necessary for successful withdrawal of
molten material. Additionally, significant preheating is required for allowing the use of corrosion
resistant materials, (silicon carbide, alumina), for dual withdrawal. The method of preheating used
in these experiments, allowing the tube to heat in the furnace before withdrawal, was chosen for
simplicity. It eliminated the need to design and construct a preheating system which was
compatible with high temperature operation in air, and gave the flexibility to work with the moving
tube design. It is desirable from an engineering reliability standpoint to use as simple a system as
possible, so evaluation of withdrawal without a purpose built preheater was performed.
For the experimental geometry used, this method was found to be ineffective. Only the lower third
of the tube actually entered the furnace during preheating so most of the tube remained cold,
leaving thermal shock a problem for the upper portion of the tube. Additionally, heating was
provided by radiant heat from the 1600 °C furnace tube and furnace insulation. In a cold walled
1
lunar MRE cell heating is provided internally by joule heating during electrolysis. Since the cell
wallsand closure are not expected to be very hot there will not be an effective means of providing
radiant heat to the tube.
The final requirement necessitating a new method is the cooling and draining process after
withdrawal. Oxide ceramics, while stable in the melt for periods long enough to facilitate many
withdrawal cycles, have significant thermal shock trouble when removed from the furnace. Also,
keeping the tube hot provides significant benefit for draining after withdrawal. It was observed that
the lower portion of the withdrawal tube which was held in the furnace to drain consistently had
little or no frozen plug inside after a two minute draining period. The upper portion, however, did
not get hot enough during withdrawal and cooled too quickly to drain completely thus creating a
plug. If the cooling rate over the entire length of the withdrawal tube was controlled both thermal
shock and freezing inside the tube could be prevented, perhaps indefinitely.
The easiest way in which the control of heating and cooling could be achieved would be by use of
a resistance coil and radiation shields. In the vacuum on the lunar surface refractory metal
resistance wire and radiation shielding could be used without concern for oxidation and would
provide a low cost and long life solution for external heating of the tube both before and after
withdrawal. The larger difficulty with this solution lies in the need to translate the withdrawal tube
vertically and the requirement that the heating/ insulation be compatible with this design
approach. Another drawback lies in the energy cost of operating the external heating equipment
which could require considerable power to reach temperature.
6.1.2 Pressurized Removal Modification
The most significant difference between the operation of the withdrawal device as tested in this
work and the lunar configuration is the pressurizing method. In this case vacuum over the mold
was used, effectively using atmospheric pressure to push the fluid out of the crucible. On the
moon the absence of an atmosphere will require the use of a sealed pressure vessel to achieve
removal of material from the MRE cell. As described in section 4.3.2 this will require the use of a
fundamentally different seal, one which is made by the translation of the tube into the sealing
position, and will preferably allow dynamic sealing. In the experiments described the vacuum
dome was sealed with a static o-ring and glue around the tube. On the moon the seal will have to
be made by contact between two sealing faces as the withdrawal tube moves into position. Prior
to withdrawal the MRE cell will have to operate in a closed container to allow capture of oxygen
from the electrolysis process. When withdrawal is required the cell will have to be opened to allow
the insertion of the withdrawal tube. This will require that the cell is either pumped empty of
atmosphere or vented, accepting loss of any contained oxygen. Once a port is opened, the tube
will drive into the cell; any frozen crust will have to be broken if necessary. Upon reaching the
1
appropriate withdrawal position the tube will have to be sealed against the cold walled pressure
vessel to allow repressurization to remove the cell contents.
It would be beneficial for control of the process to allow this withdrawal position to be flexible to
some degree. This, however, requires some consideration towards the design of the sealing
method. Several concepts are proposed below which would allow sealing between the cell
enclosure and the hot withdrawal tube while allowing an adjustable withdrawal position.
Clearance Seal
T ~1600°C
Ceramic Gasket
0 ATM
T ~1600 °C
P > 0 ATM
P> 0ATM
0 ATM
P > 0 ATM
Hi-Temp Static +
Cooled Dynamic
Seal
Static
Seal
0 ATM
T ~1600 °C
P > 0 ATM
Pressurized Withdrawal
Cooled
Dynamic
Seal
Figure 28: Schematic diagrams of potential high temperature dynamic sealing options. An
additional concept not presented is the use of an edge welded bellows to achieve a small degree of
translational freedom.
The clearance seal concept incorporates two precision fit and aligned components to restrict the
gas escape pathway to a narrow annulus. For the small pressure differential and duration
required for withdrawal this option would enable a sufficient seal for operation with little gas loss.
The benefit to this method is that it allows the elimination of wearing parts or the need for cooling.
This comes at the expense of requiring precise alignment and no interfering particulate or buildup
2
in order to prevent binding of the tube in the bushing. A tradeoff between gas loss and alignment
precision required would enable adjustment of the dimensions to an optimum point.
The ceramic gasket method employs a ceramic gasket to provide some compliance in the sealing
system. A woven gasket with light interference with the tube would enable containment of
pressure and eliminates the need for precise alignment. The drawback however is that the seal
will eventually wear. A metallic gasket could be used to reduce wear, but material choice will be
dependent on the temperature the gasket sees during withdrawal. In a high temperature oxidizing
environment a woven wire gasket will potentially fail very quickly. The best option with this
configuration may therefore be a metal reinforced ceramic gasket.
The last option pictured is a setup in which the temperature and dynamic capacity of the seal are
separated. In this case a high temperature contact seal is made against a moving plunger which
is cooled at the sealing surface during the withdrawal process. Ensuring sufficient distance
between the hot and cold sealing locations would prevent detrimental temperature drop of the
fluid during withdrawal and damage to the dynamic sealing gasket. This method requires the
added complication of a cooling means and may still be difficult given the inability to use typical
lubricants on the lunar surface. Another similar method would employ a metal bellows to allow a
static sealing surface to be translated vertically during withdrawal. This may be easier than a
cooled seal but still imposes a service life as edge welded bellows typically have limited number
of extension compression cycles. Formed bellows would not be appropriate for this purpose
because of their low compressibility, (it is desired to maintain as low a possible profile as possible
to ensure minimum withdrawal distance).
6.1.3 Automated Pressure Differential Control
Another revision to the system that needs to be addressed is control of the pressure differential
used to withdrawal fluid. It was seen in many cases that it was difficult to obtain accurate
adjustment of pressure in the vacuum dome and the result was splashing of molten material
around the chamber. Automated, rather than manual pressure control will be necessary to ensure
a repeatable process. This will enable accurate comparison of results between runs, indicating
necessary adjustments in future experimental work.
6.2 Process Modeling
In order to develop a more effective hardware configuration and optimum process parameters it
will be necessary to engage in a comprehensive process modeling effort. Modeling of heat
transfer and fluid flow will give a picture of the acceptable limiting parameters for the process and
will enable designers to predict performance before integration into electrolysis cell. While an
analytical approach to modeling coupled fluid and heat transport during the withdrawal process
3
seems out of the question, a numerical solution to the problem is likely possible. However, to
perform such modeling, a series of inputs must be given to the model: thermophysical properties
of the fluid and the tube, initial conditions for the components in the system, and the operating
parameters of the experiment, namely the time-pressure profile used during withdrawal. This
work gives some idea of the range of parameters necessary for success, and therefore provides
a starting point for modeling of the process.
Performing modeling of the withdrawal process was explored during the project, but as this
particular process has never been performed and detailed input data for a coupled numerical
model was not available, the effort to engage in this work was deemed outside the scope of the
project. Additionally, to be considered useful for design purposes any model developed would
need to first be validated with experiment. Since considerable experimental work would be
required before a model of use could be produced it was decided to simply press on with
experimental design and validation of the pressure differential withdrawal concept before time
was spent engaging in a serious modeling effort.
6.3 General Re-Design Considerations
It is apparent from the discussion above that considerable work will be required to create a
system which has functionality similar to that of a lunar compatible system. While the next stage
of experimentation can be continued with the same style of setup used for these experiments,
development of a fully functional system will required complete redesign. The system used was
designed around limiting hardware and budget constraints that prevent the most efficient solution.
Combined with a physical model of the withdrawal process and a clearer picture of the
requirements and operational configuration of a self heating cell, future designers will be able to
create an optimized withdrawal mechanism centered on the basic principle tested in this work.
4
Section 7.
Conclusions
As part of a larger project to develop a means of producing in situ oxygen on the lunar surface,
this project investigated the feasibility of a method of removing molten materials from a simulated
electrolysis cell. The investigation included the design, construction and experimental validation
of a pressure differential fluid transfer system which utilizes vacuum to pull both molten
ferrosilicon and a lunar regolith like oxide melt from a furnace at 1600 °C. The effectiveness of
this pressure differential withdrawal method was validated through a series of experiments
evaluating performance with aluminum, ferrosilicon, a regolith melt depleted of iron oxide and
partially depleted of silica, and finally a dual layered melt containing both ferrosilicon and the
depleted regolith. This work also included an investigation of materials solutions for containment
of lunar regolith both for electrolysis experimentation and for materials handling components.
The differential pressure withdrawal method was selected for removal of material from an
electrolysis cell for several reasons. It was determined that there is not a good solution for
containment of the regolith and ferrosilicon melts for any long period of time. This makes the best
option for melting a cold walled reactor in which the edges of the melt freeze, providing its own
container. The pressurized withdrawal method allows easy interface with this type of cell; it is
widely used in the aluminum industry. In addition to compatibility with a cold walled electrolysis
setup, the withdrawal method also has the potential to allow tight control over the removal
process giving operators the ability to maintain fluid levels as necessary for optimum cell
performance. Finally, properly implemented, the method is good from a reliability standpoint.
Moving parts are kept to a minimum and temperature exposure is limited to one or two critical
components. Importantly, the method also allows circumvention of the unsolved long term
materials compatibility issue.
Experimental work with the system designed has provided proof of concept of a pressure
differential materials removal system and suggests that the process can be made to work on a
larger scale for lunar use. Of the various options evaluated, the material best suited for critical
use as a withdrawal tube was hot pressed boron nitride. This material was seen to exhibit
superior thermal shock resistance during withdrawal allowing heating of nearly 1000 °C/s.
Corrosion during short term exposure was minimal and though oxidation of the tube was present
the functionality of the tube was not affected nor were dimensions altered. Results of
experiments, however, indicate that considerable work needs to be focused on preheating the
1
tube. Even after holding the tube in the furnace for draining, a frozen plug repeatedly formed in
the withdrawal tube preventing long term use. Of note, the best material evaluated for corrosion
resistance against a JSC-1A type lunar regolith melt was a 99.6% pure aluminum oxide. Under
an oxidizing atmosphere, this material was found to resist corrosion for periods exceeding 5
hours in exposure to regolith like melts. In depleted regolith melts absent of iron, this material
exhibited resistance to a two layered ferrosilicon + oxide melt for periods of up to 9 hours at 1600
°C. In a reducing atmosphere it was found that this ceramic is attacked much more aggressively
than in an oxidizing environment for melts both containing and devoid of iron oxide.
Considerable further work with this method is required for successful integration into an
electrolysis cell and implementation on a lunar scale. Three stages to this work are apparent.
First, further validation with the system hardware as it stands needs to be done. In addition to
developing a preheater, especially important is investigating the ability of the system to function
cyclically, as the equipment used in this project was used a maximum of two trials before
replacement. Second, comprehensive process modeling should be done to provide insight into
the nature of hardware configuration modifications. This work provides some data which can be
used for design and validation of first stage modeling. Finally, a complete redesign of the
hardware and will be necessary to obtain a solution which can be evaluated as a lunar compatible
system. Modeling of the withdrawal process will be very helpful for providing suggestions for
component dimensions and process parameters, and will be critical to obtaining an optimum
solution.
1
Appendix A: Experimental Geometry
Figure 29: Cross section of furnace and vacuum chamber. Several dimensions, (mm), are presented
for scale.
1
Appendix B: Additional Experimental Data
This section presents data collected from various withdrawal experiments. Given considerable
trouble with data acquisition temperature and pressure data was not accurately recorded for all
experiments. The information presented here is intended to give a representative picture of the
parameters recorded during the experiments.
Figure 30: Pressure from dual withdrawal Trial #1.
2
Figure 31: Temperature measurements from dual withdrawal Trial #4.
Figure 32: Temperature profile from dual withdrawal Trial #5.
1
Figure 33: Absolute pressure in vacuum chamber during dual withdrawal Trial #5.
2
References
1
Taylor L.A. and Carrier W.D.. ―Production of Oxygen on the Moon Which Processes are Best and Why‖:
AIAA Journal 30 [12] (1992).
2
Curreri P.A., Sen S. and Sadoway D.S, ―In Situ Resource Utilization—Molten Oxide Electrolysis‖,
MSFC Independent Research and Development Project No. 5–81, NASA/TM—2006–214600 (August
2006).
3
McKay D.S., Carter J.L, Boles W.W., Allen C.C., Allton J.J., ―JSC-1 A New Lunar Soil Simulant‖
Engineering, Construction, and Operations in Space IV, American Society of Civil Engineers p.857-866
(1994).
4
Sirk A.S. Private Correspondence. 12/9/09.
5
Hall C.M., U.S. patent 400,664 (2 April 1889).
6
Stefanescu D.M., Curreri P.A., and Sen. S., ―Molten Materials Transfer and Handling on the Lunar
Surface‖. Proceedings Space Technology and Applications International Forum 2007.
7
Papike J.J., Simon S.B., and Laul J.C. ―The Lunar Regolith: Chemistry, Mineralogy and Petrology‖, Rev.
Geophys. Space Phys., 20, pp. 761-826 (1982).
8
Lay, L.A. Corrosion Resistance of Technical Ceramics, London H.M.S.O. (1983).
9
Sundaram, S.K., J. Hsu, R. Speyer, ―Molten Glass Corrosion Resistance of Immersed CombustionHeating Tubes in Soda-Lime-Silicate Glass‖, J. American Ceramic Society 77 [6] 1613-23, (1994).
10
Sundaram, S.K., J. Hsu, R. Speyer, ―Molten Glass Corrosion Resistance of Immersed CombustionHeating Tubes in E-Glass‖, J. American Ceramic Society 78 [7] 1440-46, (1995).
11
Hunt, A. , Chinn, R., ―A Ceramographic Evaluation of Chromia Refractories Corroded by Slag‖; Office
of Science and Technical Information, USDE (2002)
12
Bennet J., Kwong K., Powell C., Thomas H., and Krabbe R., ―An Analysis of the Causes of Failure in
High Chrome Oxide Refractory Materials from Slagging Gasifiers,‖ Office of Science and Technical
Information, USDE (2005)
13
Wuchina E., Opila E., Opeka M., Fahrenholdz W., and Talmy. I., ―UHTCs: Ultra High Temperature
Ceramic Materials for Extreme Environment Applications‖ The Electrochemical Society Interface 16 [4]
pp. 30-37 (2007).
14
Hot-Pressed Boron Nitride Shapes; Momentive Performance Materials, 2007. May 23, (2008)
15
Naka M., Kubo M., Okamoto I. Wettability of Silicon Nitride by Aluminum, Copper and Silver. Journal
of Materials Science Letters [6] p.956-966 (1987).
16
Deal B., and A. Grove, J. App. Phys., 36, 3770 (1965)
17
Metallurgical Thermochemistry, Fifth Edition C.B. Alcock O. Kubaschewski, Pergamon Press, New
York, 1993
18
Kalogeropoulou S. Baud L. and Eustathopoulos N. ―Relationship Between Wettability and Reactictivity
in Fe/SiC System. Acta Metall. Mater. Vol 43, N. 3, pp. 907-912 (1995)
19
Private conversation with C. Weems, Griffin Wheel Company. 12, December 2009.
20
Private conversation with A. Tabereau, ALCOA. 9, August 2009.
21
Kondrateiv A., Hayes P. C., and Evgueni J. ―Development of a Quasi-chemical Viscosity Model for
Fully Liquid Slags in the Al2O3-CaO-FeO-MgO-SiO2 System. Part 2. A Review of the Experimental Data
and the Model Predictions for the Al2O3-Cao-MgO, CaO-MgO-SiO2 and Al2O3-MgO-SiO2 Systems‖. ISIJ
International, vol. 46 No. 3, pp. 368-374 (2006).
22
Sirk A.S. Private Correspondence. 9/2/2009.
3
23
Takamicih Iida, Roderick I. L. The Physical Properties of Liquid Metals. Guthrie Oxford Press, New
York, 1988.
24
Hull, R. Properties of Crystalline Silicon, IET, 1999.
25
Sirk A.S. Private Correspondence. 12/9/2009.
26
Assael M. J., Kakosimos K., Banish R. M., Brillo J., Efry I., Brooks, R, Quested P.N., Mills K. C.,
Nagashima A., Sato Y., and Wakeham W. A. ―Reference Data for the Density and Viscosity of Liquid
Aluminum and Liquid Iron‖, J. Phys. Chem. Ref. Data, Vol. 35, No. 1, 2006.
4