Page 1
Study of Properties of Cryolite – Lithium Fluoride Melt containing Silica
by
Sridevi Mariyam Thomas
A thesis submitted in conformity with the requirements for the degree of Master of Applied Science
Materials Science and Engineering University of Toronto
© Copyright by Sridevi Mariyam Thomas 2012
Page 2
ii
Study of Properties of Cryolite – Lithium Fluoride Melt containing
Silica
Sridevi Mariyam Thomas
Master of Applied Science
Materials Science and Engineering
University of Toronto
2012
Abstract
The ultimate goal of this study is to examine the feasibility of extracting silicon from silica
through electrolysis. The objective of the thesis was to evaluate the physico-chemical properties
of a cryolite-lithium fluoride mixture as an electrolyte for the electrolysis process. A study of
86.2wt%Cryolite and13.8wt%Lithium fluoride melt with silica concentration varying from 0-
4wt% and temperature range of 900-1000°C was done. Three properties were measured using
two sets of experiments: 1) Dissolution Behaviour Determination, to obtain a) solubility limit, b)
dissolution rate (mass transfer coefficient) and 2) density using Archimedes’ Principle. The study
concluded that solubility and dissolution rate increases with temperature and the addition of LiF
to cryolite decreases the solubility limit but increases the rate at which silica dissolves into the
melt. With addition of silica, the apparent density of electrolyte first increases up to 2-3wt% and
the drops.
Page 3
iii
Acknowledgments
My sincerest thanks go to Professor Barati for giving me the opportunity of working on this
thesis and his invaluable guidance and support. I am most grateful for the freedom with which he
allowed me to work on this project letting me learn a lot more about my subject than I would
have otherwise done.
I would like to thank the Materials Science and Engineering Department for the financial
assistance and the close-knit community within the department for their support. My gratitude to
Professors Argyropoulos and Utigard as well for first giving me the opportunity to work on
metallurgy related projects in my undergraduate studies. I would also like to thank Sal Boccia of
the MSE department and George Kretchmann of the Geology department for their advice and
assistance in various analyses.
I am indebted to my parents for pushing me all my life to achieve my best and doing whatever
they could to aid me. I am thankful for the support they gave me when I made the most important
decision to date; choosing to switch to materials science. Had I not, I would not have discovered
my love for metallurgy and for this I am always grateful.
I really appreciate the support of the Sustainable Materials Processing Research Group especially
Karim and Samira. A lot of the work done for this research would not have been possible without
Karim’s advice and Samira’s help.
Finally, my express thanks to my friend Michal Fulmyk, for his criticisms and lending ear to the
many discussions about the experimental procedure and results obtained. His insight as an
engineer of a different discipline helped me shape this thesis.
Page 4
iv
Table of Contents
Acknowledgments .......................................................................................................................... iii
Table of Contents ........................................................................................................................... iv
List of Tables ................................................................................................................................. vi
List of Figures ............................................................................................................................... vii
List of Abbreviations and Symbols ................................................................................................ ix
Chapter 1 Introduction .................................................................................................................... 1
Chapter 2 Literature Review ........................................................................................................... 5
2.1 Solar Power and the Role of Silicon ................................................................................... 5
2.2 Current Methods of Extraction and Purification of Silicon .............................................. 10
2.2.1 Siemens Process .................................................................................................... 11
2.2.2 Zone Refining ....................................................................................................... 13
2.2.3 Other Methods ...................................................................................................... 14
2.3 Generation of Si by Electrolysis ....................................................................................... 17
2.4 Operating Temperatures .................................................................................................... 20
2.5 Choice of Electrolyte ........................................................................................................ 24
2.6 Measuring Properties of Electrolyte ................................................................................. 25
2.6.1 Determining Dissolution Behaviour ..................................................................... 25
2.6.2 Measuring density ................................................................................................. 30
Chapter 3 Experimental ................................................................................................................ 34
3.1 Dissolution behaviour ....................................................................................................... 35
3.1.1 Examined Approaches .......................................................................................... 35
3.1.2 Selected Method .................................................................................................... 41
3.2 Density Measurement ....................................................................................................... 47
3.3 Conductivity Measurement ............................................................................................... 51
Chapter 4 Results and Discussions ............................................................................................... 56
4.1 Dissolution Behaviour ...................................................................................................... 56
4.2 Density .............................................................................................................................. 62
Chapter 5 Conclusions and Future Work ...................................................................................... 69
5.1 Summary and Conclusions ............................................................................................... 69
5.2 Future Works .................................................................................................................... 71
Page 5
v
References ..................................................................................................................................... 72
Appendices .................................................................................................................................... 78
Appendix A – Phase Diagrams generated by FactSageTM
....................................................... 79
Appendix B – Derivation of Equation 7 .................................................................................. 84
Appendix C – Effect of considering V and h as constants on k and CS ................................... 89
Appendix D – XRD analysis of white deposit on silica rod in dissolution behaviour
experiment ................................................................................................................................ 91
Page 6
vi
List of Tables
Table 1 Impurity ranges for MG-Si and 11.5% efficient Si solar cell [9] ...................................... 8
Table 2 Material requirements for an electric arc furnace producing silica [10] .......................... 11
Table 3 Typical analysis of acid leached MG-Si [9] .................................................................... 15
Table 4 Voltages required for electrolysis of Ta2O5 and Nb2O5 [6] ............................................. 18
Table 5 Impurity levels in ppm for different silicon sources and products [26] ........................... 19
Table 6 Additive effect on cryolite-alumina electrolyte [33] ....................................................... 22
Table 7 Conditions of various studies on electrolysis of silicon .................................................. 32
Table 8 Effect of time on the calculated k and Cs ........................................................................ 59
Table 9 Molar volumes of the various constituents of the melt .................................................... 65
Page 7
vii
List of Figures
Figure 1 Price make-up of solar cells [3] ........................................................................................ 2
Figure 2 Percentage contributions from each energy source [8] .................................................... 6
Figure 3 Estimates of available energy resources using today's technology [8]............................. 6
Figure 4 Efficiency drop with impurity content [2] ........................................................................ 8
Figure 5 Correlation between purity and cost of silicon [9] ........................................................... 9
Figure 6 Schematic representation of the Siemens process [1] .................................................... 12
Figure 7 Alumina solubility versus cryolite ratio ......................................................................... 23
Figure 8 Schematic for measuring diffusion coefficients by rotating disc method [51] .............. 29
Figure 9 Modified schematic to observe dissolution behaviour for this study ............................. 29
Figure 10 Density of the electrolyte with alumina and silica additions to cryolite [45] ............... 30
Figure 11 a) SEM Image of a 6wt% sample b) Si EDS of sample ............................................... 36
Figure 12 Sample size of 2g containing 6 wt% silica ................................................................... 36
Figure 13 Anticipated diffusion profile across a silica tube with CR+LiF melt ........................... 38
Figure 14 Diffusion Experiment Sample ...................................................................................... 39
Figure 15 SiO2 profile across a 2mm tube held in CR+LiF melt for 30 minutes ......................... 39
Figure 16 Effect of convection flow on the concentration profile ................................................ 40
Figure 17 Capillary reservoir to measure diffusivity .................................................................... 40
Figure 18 Schematic of the experimental setup for dissolution behaviour study ......................... 42
Figure 19 Rods after the experiment ............................................................................................. 43
Figure 20 Immersed portion of rod ............................................................................................... 43
Figure 21 Reproducibility of dissolution behaviour experiment .................................................. 45
Figure 22 Set up of the sink to be immersed in the CR+LiF melt. ............................................... 49
Figure 23 Schematic of the experimental setup for density measurement ................................... 50
Figure 24 Schematic of the experimental setup for conductivity measurement ........................... 51
Figure 25 Typical Nyquist plot for impedance spectroscopy ....................................................... 52
Figure 26 Two arrangements for four probe electrode ................................................................. 53
Figure 27 Comparing results from different experimental setup .................................................. 55
Figure 28 Typical mass loss versus time graph (1000C) ............................................................ 56
Figure 29 k versus temperature ..................................................................................................... 57
Figure 30 CS versus temperature ................................................................................................... 58
Page 8
viii
Figure 31 Comparison of variation of mass loss for different temperatures ................................ 61
Figure 32 Density versus silica concentration for five temperatures ............................................ 63
Figure 33 Density comparison with and without LiF at ~1000°C ................................................ 64
Figure 34 Density comparison of silica versus alumina addition at 1000°C ................................ 64
Figure 35 Comparison of effect of LiF addition with different oxides ......................................... 65
Figure 36 Density versus temperature for 1wt% SiO2 electrolyte ................................................ 66
Figure 37 Variations of density with silica concentration versus temperature ............................. 67
Figure 38 Density compared to liquid aluminium ........................................................................ 68
Page 9
ix
List of Abbreviations and Symbols
A area (cm2)
CB bulk concentration (g/cm3)
CR+LiF cryolite and lithium fluoride
CS saturation concentration (g/cm3)
D diffusivity
DC direct current
EDS energy-dispersive X-ray spectroscopy
EG-Si electronic grade silicon
EIS electrochemical impedance spectroscopy
EPMA electron probe microanalyzer
ICP inductively coupled plasma
k mass transfer coefficient (cm/h)
m mass (g)
MG-Si metallurgical grade silicon
ppma parts per million atoms
PV photovoltaic
SEM scanning electron microscope
SoG-Si solar grade silicon
t time (h)
XRD X-ray diffraction
XRF X-ray fluorescence
ΔG Gibb's free energy
ΔT change in temperature
Δρ change in density
#N number of nines of purity e.g. 2N = 0.99
Page 10
1
Chapter 1 Introduction
This thesis is in aid of exploring the feasibility of electrowinning solar grade silicon (henceforth
referred to as SoG-Si) from a cryolite, lithium fluoride and silicon dioxide electrolyte bath.
Hence the purpose of this research was to conduct experiments for measuring properties of a
cryolite and lithium fluoride mixture (hereafter referred to as CR+LiF) with varying silica
concentrations and at temperatures between 900 to 1000°C. The end goal is to obtain silicon of
solar grade purity for solar cell applications.
Solar energy is available in most inhabited areas of the earth on a daily basis, more accessible
than hydro and geothermal sources and more reliable than wind and tidal as it is more consistent.
The technology for harnessing this energy is well developed by using solar cells which is
typically made of silicon, usually 10-20% efficient. Over 90% of the today’s photovoltaic (PV)
systems use silicon as the conversion material [2]. There are four main reasons why silicon is the
preferred material for solar cells:
1. Abundance: common in its oxide form quartzite SiO2, silicon constitutes 25.7% of the earth’s
crust by mass [2]. This makes it relatively cheap compared to other semiconductor materials.
2. Non-toxicity: silicon is inherently stable. However the current method of purification
(Siemens method) is toxic and efforts are underway to find a different route to extract and
purify silicon.
3. Acceptable band gap: silicon has a band gap of 1.11 eV which is very close to the optimum
band gap of 1.4eV [2]. The optimised value is determined by a balance between the electrical
energy created; a function of voltage created when electrons are excited and energy lost; a
function of current of the same. Gallium (III) arsenide, as another semiconductor material has
a more favourable band gap (1.43eV) but is not as abundant as silicon and is more toxic.
4. Good conversion efficiency: usually >10% using a simple device; the solar cell. Solar cell
devices are projected to have a 30 year operating life and their manufacture is a mature
technology.
Pertinent to the development of the solar grade silicon is a balance between its purity and cost
both of which are determined by the extraction and purification process. Although numerous
methods have been proposed for the generation of SoG-Si the problem appears to be low
Page 11
2
production rate and high cost of the product. The economy of solar power depends on 1) the
making of the device and the technology involved 2) the efficiency of the product and 3) the
extraction method of material from ore. The cost of material is primarily dictated by the
production rate and energy consumption of the extraction processes involved. It has been
estimated that the silicon material accounts for 45% of the final price of the solar cell [3]. Figure
1 shows the proportions of cost with respect to semiconductor material, processing of the module
and the assembly of the module [3]. It implies that a reduction in the materials cost results in
substantial decrease in the device price. Therefore, a process capable of producing low cost SoG-
Si will contribute to lower cost PV electricity.
Figure 1 Price make-up of solar cells [3]
The current technology for generation of high grade silicon is a two stage process 1) extraction
of crude silicon through carbothermic reduction of silica and 2) purification of silicon involving
chemical vapour deposition (Siemens’ process) followed by directional solidification. The
method is essentially similar to the process of making electronic grade (semiconductor) silicon
that requires much higher purity (9N) than that of SoG-Si (6-8N). The disadvantages of such
method are low production rate and high energy consumption which lead to prohibitive costs for
solar uses. The ultimate goal of this project is to investigate the possibility of using electrolysis
as a less studied but promising technique for producing low-cost SoG-Si. Electrolysis has the
potential to combine extraction and purification into a one step process to obtain solar grade or
near solar grade purity.
Electrolysis has been explored as a feasible process because of the similarity of SiO2 to Al2O3
and Si to Al. Alumina is reduced by electrolysis in cryolite at 1000°C in liquid state (voltage of
Module Assembly
24%
Process 31%
Silicon 45%
Page 12
3
4V, current density 1A/cm2, efficiency 0.9) [4] to produce nearly all the aluminium that is made
today. The resulting aluminium is denser than the electrolyte so it sinks to the bottom of the bath
and is prevented from oxidation by the atmosphere. Al ion requires 3 electrons to be reduced to
metal, while Si which is next to Al on the periodic table, needs 4 electrons. This higher energy
requirement is offset due to silica being available in high purity at a relatively low cost (unlike
alumina which requires the Bayer process for refining bauxite) making the end product a
comparable price to aluminium.
Purity of the silicon plays an important part as impurities will reduce the efficiency of the solar
cells. Purification inherent to electrolysis of Si is carried out due to the different deposition
voltage of elements. Elements with a lower deposition voltage than Si will deposit out first so a
low voltage can be applied to sacrificially remove them. Elements with higher deposition voltage
will stay in the solution. This means the electrolyte will need to be purified (using electrolysis)
after a while to prevent build up else these impurities will deposit out due to saturation of the
solution [4-6].
Carbothermic reduction involves an arc furnace and a carbon source all which introduces
impurities into the silicon making it only 95-99% pure. The silicon that is obtained from
electrolysis has been shown to have a higher purity than MG-Si and so can be used as a precursor
to zone refining. For instance, in zone refining the number of repetitions can be reduced if
sufficient purity of Si has been achieved by electrodeposition.
Carbothermic reduction uses non-renewable coke and results in the emission of CO2. Electrolysis
uses electricity which can be obtained from a clean and/or renewable resource such as hydro or
solar, resulting in a process with a very small CO2 footprint compared to today’s technologies.
The reason the electrolysis of silicon is not as of yet applied on an industrial scale is because of
the high melting temperature of silicon, 1414°C, which means to keep it molten, high operating
temperatures must be maintained. High temperature electrolysis will be accompanied by such
problems as stability of the electrolyte and choice of electrodes and the furnace walls. Lowering
the temperature would assist solving these problems and also reduce energy consumption but this
would mean the silicon is deposited as a solid. Liquid deposition is favoured to solid because 1)
inclusions of impurity particles are reduced due to the liquid surface tension and 2) since solid-
liquid interface energy does not need to be overcome like in solid deposition where increasing
Page 13
4
the current density to increase deposition rate leads to powdery solid. The recovery of such
powdery product from electrolyte will also be difficult and involve significant losses. The
method proposed by our group is to overcome both problems by depositing the silicon in a liquid
metal cathode resulting in an alloy of low melting temperature, thereby reducing operating
temperatures and increasing deposition rate. The alloy is then slow cooled to precipitate out
almost pure silicon dendrites from the alloy that are then removed by physical or chemical
separation methods. In addition to lower temperature, the method has the advantage of producing
an alloy heavier than Si that settles to the bottom and is protected against oxidation.
In order to identify a suitable electrolyte and characterise it, this fundamental study was carried
out. The literature review indicated that cryolite was preferred among previous authors. To lower
the working temperature, lithium fluoride was added to reduce the liquidus temperature. Three
properties were investigated between 900 and 1000°C:
1. Solubility limit of silica in the CR+LiF electrolyte.
2. Dissolution rate of silica in the CR+LiF electrolyte by calculating a mass transfer
coefficient.
3. Density of the CR+LiF electrolyte with varying silica concentrations between 0-4wt%.
Having knowledge of these (and other) properties is essential in determining whether electrolysis
of the CR+LiF melt containing silica to obtain silicon in its pure or alloy form is a viable option.
Chapter 2 of this thesis will provide an overview of the previous works in this area; the various
types of electrolytes, the temperature of deposition; liquid silicon versus solid silicon deposition
and the additives used to improve electrolyte properties. Since the aim of this thesis is to look at
the properties of CR+LiF bath the literature review will also cover techniques on how to measure
solubility, mass transfer coefficient and density. The review will explore the viability of using
electrolysis of silica to extract silicon and explain why this thesis has chosen cryolite as the base
for the electrolyte. Chapter 3 on materials and methods will describe the experimental procedures
used and the way the data was collected and processed to obtain said properties. Chapter 4 on
results and discussion will examine the results and the conclusions that may be drawn from the
obtained data and how this helps to realise the method in an industrial setting. Finally, the major
conclusions of the study will be presented in Chapter 5.
Page 14
5
Chapter 2 Literature Review
2.1 Solar Power and the Role of Silicon
A renewable source of energy is an absolute must because of our world’s dependency on non-
renewable sources in the form of chemical energy i.e. fossil fuel resources; which as the name
suggests cannot be replaced, certainly not at the current rate of consumption. At the present rate
of usage, it is estimated that the earth resources will be depleted of oil in 46 years, natural gas in
100 years and coal in 168 years [7]. Fossil fuels also have a further disadvantage in that their use
creates greenhouse gases (GHG), mainly CO2, which are thought to be the primary reason for
global warming. Recording of the average temperature over ten-year periods have shown that the
mean temperature has increased over the decades.
Current options for renewable resources are solar, hydro, wind, biomass and geothermal energy.
Technology to capture tidal and wave energy are still in the infant stage but the technologies for
the other sources have grown due to concern about climate change and the depletion of fossil
fuels. About a fifth of the world’s electricity supply is already generated by renewable energy as
shown in Figure 2 [8] most of which is provided by hydro power. Hydro has the advantage that it
has a high energy density per volume but as observed in Figure 3 [8] when compared to the
world energy use hydro cannot meet the demand. While wind and geothermal energy can
potentially meet current world energy use, this may not be the case in the future since as
population and industrial growth continues, energy use increases. Solar energy, on the other
hand, can more than 3 times fulfill the current energy need which is why it is important to
develop technologies for harnessing this energy source.
Page 15
6
Figure 2 Percentage contributions from each energy source [8]
Figure 3 Estimates of available energy resources using today's technology [8]
Coal/Peat 40.3
Gas 21.4
Oil 5.1
Nuclear 13.4
Hydro 16.5
Solar 0.1 Geothermal 0.3
Wind 1.4
Biofuels 1.4
Others 0.1
Other
Renewables 3.3
477
>1600
600 500
>250
50 <1
0
500
1000
1500
2000
En
ergy F
low
(ex
ajo
ule
s p
er y
ear)
World Solar Wind Geo- Biomass Hydro- Ocean
Energy thermal power
Use
Page 16
7
Current global consumption of energy is 16.3TW. Energy available from sun is estimated at
86,000TW. With a typical 12% efficient solar cell at least 136TW of the sun’s energy need to be
captured to satisfy the world’s energy demand. On average an area of 1m2 can produce 100W of
DC power in direct sunlight (if efficiency is 12%) [2]. To produce 136TW an area of
1,360,000km2, about a seventh of the land area of Canada, is required. Theoretically it is possible
to implement a totally solar power dependent planet with storage methods implemented.
However it may not be practical due to the transient nature of solar power, the need for chemical
energy in many application, and cost. Nevertheless there is a consensus that it can play a big role
in our energy supply.
A solar cell is a device that absorbs solar energy and uses it to promote an electron from the
valence band to the conduction band of the semiconductor material used (leaving a hole in the
valence band) and conducts electricity. Solar radiation is a spectrum of electromagnetic waves
(EM waves) of different energies. The percentage of solar energy that is utilised depends on the
energy gap between the conduction and valence bands. Only EM waves with energies above the
energy of the gap can be absorbed to promote the electron and any extraneous energy is used to
heat up the cell material.
As discussed in Chapter 1, silicon is a suitable choice for semiconductors in solar cells because it
has a band gap close to the optimal band gap that would make maximum use of the EM waves
with minimal energy loss. To do this at an efficiency of 10% the silicon used in solar cells need
at least a purity of 6N but it is crucial to remove the dopant impurities such as P, B and As which
degrade the solar cell performance.
The purity of metallurgical grade silicon, obtained from carbothermic reduction of silicon
dioxide, usually only reaches a maximum of 2N. Impurity atoms can act as electron scatterers
that impede the mobility of electrons; they lose energy and fall back into the valence band and
recombine with a hole which results in the destruction of charge carriers. Figure 4 shows the
efficiency drop off with impurity content for silicon cells [2]. Table 1 shows impurity ranges for
certain elements in MG-Si given in the first column. The second column shows how much they
need to be reduced to make solar cells of 11.5% efficiency (to make 100W/m2). Also included
are segregation coefficients which are important for zone refining [9].
Page 17
8
Table 1 Impurity ranges for MG-Si and 11.5% efficient Si solar cell [9]
Impurity Element Conc. Range in
MG-Si (ppm)
Conc. required for 11.5%
efficient Si solar cell (ppm)
Segregation
coefficient
Al 1000-4000 25 2.0 x 10-3
B 40-60 0.01 0.8
P 20-45 0.01 0.35
Cr 40-220 <5 1.1 x 10-5
Fe 1500-6000 <5 8.0 x 10-6
Cu 15-40 <5 4.0 x 10-4
Mn 10-80 <5 1.3 x 10-5
Ni 10-95 <5 1.0 x 10-4
Ti 120-275 <5 2.0 x 10-6
V 50-250 <5 4.0 x 10-6
C 1000-3000 <5 0.05
Zr 15-25 <5 1.6 x 10-8
Figure 4 Efficiency drop with impurity content [2]
Page 18
9
The hardest impurities to remove are phosphorus and boron because of their relatively high
solubility limit and segregation coefficient in silicon. It is essential that they be removed since
these elements are used for doping the Si to make a semiconductor. Since overdoping can lead to
a degenerate semiconductor (essentially a useless device which conducts electricity without
promoting electrons from valence to conduction band) the dopant impurities need to be
maintained at a low and well controlled concentration. Consequently it is necessary to implement
a purification process, such as the Siemens, after the initial extraction process of carbothermic
reduction. The Siemens method, which is currently used in industry, is very time consuming and
energy intensive driving the cost of the product up.
As Figure 5 depicts, there is an inverse relationship between the impurity content and price for
various grades of silicon. Due to the lower purity requirement for SoG-Si, there seems to be a
window of opportunity to generate this material with a cost of <$20/kg. It has been discussed
that this in fact matches a target price of $10-15/kg that is believed to enable solar power, that is,
grid parity with fossil fuel and nuclear based electricity. In the past few years the market price of
SoG-Si has been fluctuating between $25 and $400/kg depending on the supply-and-demand
situation. Achieving a price in the range of $10-15/kg will make solar power mainstream which
will be when it can be a major contributor to the energy supply.
Figure 5 Correlation between purity and cost of silicon [9]
Alloy grade
Metallurigical
grade
Semiconductor
grade
Detector grade
1.E-11
1.E-10
1.E-09
1.E-08
1.E-07
1.E-06
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
0.1 1 10 100
Imp
uri
ty C
on
ten
t
Silicon Cost ($/kg)
SOLAR
GRADE
Page 19
10
2.2 Current Methods of Extraction and Purification of Silicon
The main source of silicon is from the mineral quartzite SiO2 which exists in very high purities
and has less than 1% metallic impurities. The other source is as a by-product of phosphate
fertiliser production: fluorosilicic acid H2SiF6 which can come with a purity of 98-99%. It can be
neutralised to form K2SiF6 which can be used in electrolysis to extract Si. A third and recently
looked at source is rice hulls which contain 95-99% pure silica. The US production from silica
content in rice hulls alone could supply 100,000 tons of Si/year. [4]
Crude Si is primarily produced by carbothermic reduction of silica SiO2 at 1900-2100°C in an
electric arc furnace. The product is known as metallurgical grade silicon (MG-Si) with purity
ranging from 95% to 99% [9]. The overall reaction is as follows:
SiO2 (l) + 2C (s) = Si (l) + 2CO (g) (1)
However, there are a lot of side reactions that happen at the high temperatures and side products
are made such as gaseous SiO, which react with O2 in the air to make fine dust particles of SiO2,
which along with SiC (from reactions with the graphite electrode) clog up the furnace. Apart
from loss of yield, handling these unwanted materials is also an issue. The CO gas in Reaction
(1) is burnt to form CO2 to partially recover chemical energy and the gaseous products are
currently not collected to be sequestered. The carbon for reduction is obtained from coke or coal
and wood chips which is the main source of impurities. About 1 million tons of MG-Si is
produced each year using an electric arc furnace. Typical operating parameters for this process
are shown in Table 2 [10]; for 1 ton of silicon produced, about 4 tons of non-renewable carbon is
required in the carbothermic reduction reaction. This will result in a very large CO2 footprint for
the process. The advantage of this method, however, is that production of silicon is fast and the
liquid silicon is easily siphoned off and cast into ingots.
Page 20
11
Table 2 Material requirements for an electric arc furnace producing silica [10]
Requirement Quantity
Quartz 2.9-3.1 ton
Coke 1.2-1.4 ton
Charcoal 1.7-2.5 ton
Graphite 0.12-0.14 ton (electrode consumption)
Electricity 12.5-14 MWh
= 1 ton of Silicon
To help improve the end product purity level, some authors suggest starting with purer starter
materials (such as quartz deposits from Arkansas, USA and BC, Canada with B<3ppm,
P<10ppm, Al<40ppm, and Fe<120ppm [9]) and carbon black as reductant. However this
approach may drive the cost prohibitively high, and still not yield a Si with high purity due to
contamination from electrodes, refractory, and other parts of the equipment.
Two main methods are used in industry to improve purity of the Si in photovoltaic and solar
applications: 1) Siemens’ process and 2) Zone refining, however alternatives have been
investigated.
2.2.1 Siemens Process
The first purification step for MG-Si is by the Siemens’ method. MG-Si is reacted with hydrogen
chloride gas to form volatile chlorosilanes:
Si (impure) + 4HCl (g) = SiCl4 (g) + 2H2(g) (2)
Or trichlorosilanes:
Si (impure) + 4HCl (g) = SiHCl3 (g) + H2(g) + ½ Cl2(g) (3)
The impurity chlorides can then be separated by fractional distillation due to their different
boiling points. The chlorosilanes are then reduced with hydrogen and Si atoms deposit on a pure
Si rod at 1150°C [1, 2, 9]:
SiHCl3 (pure) + H2 (g) = Si (pure) + 3HCl(g) (4)
The schematic in Figure 6 shows a simplified flowsheet of the Siemens’ process; the different
stages, the inputs and outputs [1].
Page 21
12
A high temperature is required to make sure the chlorosilane decomposes onto the pure silicon
rod (EG-Si) rather than the other parts of the chamber. This is achieved by electrically heating
the Si rods to serve as the reaction site, while keeping the chamber walls at low temperature. The
silicon made from this method is polycrystalline and the deposition rate is very slow; to make a
ton of silicon, 12 Siemens reactors need to be operated for 10 days [1]. The other disadvantage to
this method is the use of SiHCl3 and H2 gas which are explosive. Furthermore AsH3 and PH3 are
collected as by products which are poisonous gases, and HCl is produced which is very corrosive
as a gas, thus handling and disposing of them in a safe manner is very difficult. The advantage of
the method, however, is that it produces 9N silicon which has purity higher than the SoG-Si
requirements.
Figure 6 Schematic representation of the Siemens process [1]
HCl MG-Si
SiHCl3
Energy Source
H2
SiHCl3 Feed
Fractional Distillation
Vent
EG-Si
Chemical Vapour Deposition Chamber
Fluidised Bed Reactor
Page 22
13
2.2.2 Zone Refining
Zone refining involves heating a silicon rod from one end with a heating element so that the
silicon melts and slowly moving the heating zone along the rod. The melted portion of the rod
resolidifies to make a purer form of the silicon as the heating element moves away while
segregating impurities in the liquid. This is possible due to phase equilibria; the impurities tend
to stay in liquid rather than solidify into the Si crystal.
The segregation coefficient k quantifies how much of an impurity can be isolated out of the
crystal.
k = CS / CL (5)
CS is the concentration of the impurity in the solid silicon and CL is the concentration in the
liquid silicon. As can be seen from Equation (5) the lower the segregation coefficient the more
likely the impurity will stay in liquid. The last portion of the rod with all the impurities is
recycled.
Zone refining has a disadvantage in that it is a slow process that needs to be repeated several
times if EG-Si is to be produced. The heating element moves at a rate of 0.1 to 10mm per minute
to create polycrystalline silicon and 0.1 to 10mm per hour to create monocrystalline silicon. The
process is usually repeated about 10-15 times [1] and is only efficient if the segregation
coefficient of the impurity is low enough (10-5
to 10-1
). However boron and phosphorus have
segregation close to 1 and so are very hard to remove [2, 11].
Another disadvantage is that only small rods can be used; approximately 10cm in diameter. But
the purity level is much higher (with resistivity of 100Ωm) than silicon produced via directional
solidification (Czochralski method). Furthermore a monocrystalline silicon boule can be formed
through zone refining; this is important for semiconductors because grain boundaries,
microcracks and dislocations (defects in crystals) scatter electrons reducing efficiency.
To address the problem of production volume an alternative method is employed which creates
monocrystalline silicon, the Czochralski method. It requires a single crystal seed which is
lowered onto the surface of liquid silicon and pulled upwards in a controlled manner while
allowing silicon to solidify onto it. This method allows much larger boules to be created with 10-
15cm diameter, 1-2m length, 50-100kg. About 80% of monocrystalline silicon is grown via
Page 23
14
directional solidification [1]. On the other hand, the downside is that this method produces
silicon of a lower purity (resistivity of only 20Ωm) compared to zone refined silicon because a
silica crucible is required thereby allowing introduction of oxygen impurity when silica dissolves
into the liquid silicon [1].
There are other minor methods that have been looked at which are not as time consuming or
difficult as Zone and Siemens’ process. However they are either energetically expensive or not as
effective at removing crucial impurities. Section 2.2.3 describes several methods that have
achieved some success at removing specific impurities.
2.2.3 Other Methods
1. Acid Leaching [9]
This is a purification process that digests impurity precipitates. Impurities with low segregation
coefficients tend to precipitate at grain boundaries making the area near the grain boundaries
more brittle. If the polycrystalline silicon is crushed the cracks will tend to form along the grain
boundaries exposing the impurities. To ensure most of the impurities are leached away the
silicon has to be crushed to a size less than or equal to its grain size and then the exposed
impurities can react with an acid. Silicon is very stable in most acids except a combination of
hydrofluoric acid and an oxidising acid so the silicon remains unreacted while the impurity
products react and stay in solution. Mineral acids such as HCl, HF, H2SO4 and HNO3 are
normally used in a combination to leach out most if not all the different impurity precipitates.
Table 3 shows a typical analysis of crushed MG-Si before and after leaching with aqua regia
(HCl and HNO3 mix), HCl and HF at an elevated temperature to get total impurity level less than
400ppm [9].
A purity of up to 4N can be achieved but this can vary depending on the particle size the silicon
has been crushed to, the leaching temperature and length of time the leaching was done. Another
factor to take into consideration is the impurity atoms that get trapped in the crystal structure as
opposed to precipitating to the grain boundaries which cannot be leached out. This method has
the disadvantage in that crushing requires time and mechanical energy and material handling
issues arises when dealing with powdery material. It is also not very successful at removing B, P
and Cu and so cannot be used alone for purification.
Page 24
15
Table 3 Typical analysis of acid leached MG-Si [9]
Impurity Impurity Concentration (ppm)
Removal (%) Before After
Al 2080 150 93
Cr 180 5 97
Fe 2310 60 97
Mn 180 5 97
Ni 80 5 94
Ti 200 10 95
V 180 6 97
B 20 14 30
Cu 22 13 40
P 50 40 20
2. Reactive gas blowing [9, 11]
This method requires gases of Cl2, O2, SiCl4, wet H2 and CO2 (or combinations of) to be bubbled
through molten silicon resulting in the reaction of the impurities to make oxides or chlorides that
either bubble off because of volatility at high temperatures (>1450°C) or collect as dross on top
of the liquid silicon. This method is effective at removing Al, Ca, C, Mg, B, P and Ti because
they are more thermodynamically likely to react with the gasses than silicon. Silicon losses are
minimal and a purity level of at least 99.99% can be achieved. However other elements like Cr,
V, Cu, Mn, Ni and Fe will stay in the silicon and cannot be removed without sacrificing a
substantial amount of silicon since oxides and chlorides for those impurities are
thermodynamically less favourable than for silicon.
3. Slagging [9]
This process involves adding a variety of oxides (CaCO3, BaO, MgO, Al/SiO2, CaO/SiO2 and
CaF2/SiO2) to create slags which will favourably dissolve the impurity but not the silicon and
will not dissolve in the silicon. It has been shown that this method can remove Ti, Mn, V and Al
by one order of magnitude.
Page 25
16
4. Solvent Refining [11-14]
This method uses a “getter” metal to alloy with the silicon to favourably dissolve impurities.
Some metals have such low solubility in silicon that when trying to alloy it with silicon the end
product simply forms two phases; one of metal-Si alloy and one of pure silicon. If this metal has
a more favourable solubility for impurities than silicon then it is referred to as a getter metal and
the impurities leave the silicon phase to be dissolved in the metal phase. A few getter metals that
have been researched are aluminium, copper, iron and tin. The alloy can then be cooled and
crushed to separate silicon from the alloy by chemical or physical separation methods.
5. Plasma Refining [9]
This method requires the silicon be in the molten state. Electromagnetic stirring transports the
impurities to the surface of the molten silicon and a plasma torch consisting of oxygen and/or
hydrogen is blown at the surface allowing the gases in the plasma to react with the impurities.
The gases are effective at removing boron but the method is very energy intensive.
Page 26
17
2.3 Generation of Si by Electrolysis
The silicon that is deposited by electrolysis can have three microstructures; monocrystalline,
polycrystalline or amorphous. Amorphous silicon is usually deposited using organic electrolytes
and typically has a small amount of hydrogen dissolved in the silicon [15-17], though this can be
released by heat treatment. Amorphous silicon is usually used for coating for wear resistance.
They also have applications in solar cells for calculators and other devices that do not require
much power, but generally it has a lower efficiency than crystalline silicon [17]. Monocrystalline
silicon growth is more time consuming to grow because of low current density to prevent the
nucleation of new grains. In most cases, for solid silicon deposition, polycrystalline silicon can
be expected.
Cathodes used in silicon deposition can determine the macrostructure which can vary from
powdery to dendritic to coherent layers. Monnier et al. [18] managed to deposit solid silicon of
99.99% purity using a graphite cathode but the silicon tended to react with the graphite and
disintegrate it and the rest of the silicon was in a powdery form which broke off and dispersed in
the electrolyte and hence was difficult to collect. Work done by Elwell et al. [19-22] using Flinak
(eutectic LiF/NaF/KF) as the electrolyte determined that silver had minimal reaction with the
deposited solid silicon at 750°C. After adjusting operating parameters such as concentration of
the silicon source (K2SiF6), current density and deposition voltage, they finally succeeded in
getting coherent growth of silicon layers on graphite as well. Pure silicon in solid or liquid form
can have unwanted reactions with the cathode material. Sometimes however the reactions are
deliberate if an alloy or a coating is the intended end product. In either case cathodes need to be
chosen carefully.
Purification of the silicon obtained, for the purpose of this thesis, would have to ensure the end
product can be used in solar cells. Solar cells, as mentioned before, require a purity of 6N or
higher [9] so it is essential the silicon be deposited with high purity or post processing succeed in
lowering impurity levels. So far there are Siemens process and zone refining in place for post
processing MG-Si but since Siemens process is energy intensive and involves toxic gases and
zone refining is time consuming it would be preferable if a low cost processing method was
available.
Page 27
18
In electrolytic production of metals, nobility of the ions determine the voltage at which the ions
can deposit out and this is the main principle applied in “automatic” purification in the process.
This is also the reason why aqueous solutions with Si4+
ions cannot be used because hydrogen
will be produced at the cathode instead. This can already be observed to some extent in
electrolysis of organic solutions as mentioned above. Hence most of the electrolysis experiments
are done with molten salts instead. It is possible to separate out ions even if the difference in
theoretical deposition voltages between them is quite small [23, 24] as shown in Table 4. The
difference between the experimental voltages that are required to decompose and deposit
tantalum and niobium is only 0.27V.
Table 4 Voltages required for electrolysis of Ta2O5 and Nb2O5 [6]
Oxide Experimental
(V)
Theoretical without anodic
phenomenon (V)
Theoretical with anodic
phenomenon (V)
Ta2O5 0.91 1.55 0.45
Nb2O5 0.64 1.40 0.30
In electrolysis a “pre-electrolysis” processing can be employed as Elwell et al. [19, 21] have
done, which runs the cell in the same manner as one would with electrolysis but with a much
higher deposition potential and current density to deposit out the impurities with a lower
deposition potential than silicon. Some silicon will be sacrificially lost in this process. It should
be noted that in aluminium electrolysis where the electrolyte is cryolite, Na3AlF6, and alumina,
Al2O3, the deposition voltage of Na+ ions is less than Al
3+ ions. Na
+ ions do not however deposit
out first; the reason for this is yet unknown [6]. Theoretically the impurities with higher
potential than the one used in pre-electrolysis should stay in solution when depositing silicon but
this is not what is observed in practice. With co-deposition of the higher potential impurities with
silicon the Gibb’s free energy, ΔG, of the system is reduced [19]. Even so the researchers
managed to get a purity of 99.98-99.999%. Even with electrorefining similar principles can be
applied; the impurity ions that are more electronegative remain at the anode while those more
electropositive dissolve and stay in solution but do not deposit at the cathode until their
concentration reaches a certain level in the solution (then the electrolyte needs to be “cleaned”).
This is the principle used when producing copper of 99.99% purity using electrolytic refining.
An electrolyte material must be carefully chosen to avoid ions that co-deposit with silicon.
Sometimes a starting material with as high purity as possible is utilised, for instance Elwell et al.
Page 28
19
started off with electrolyte and potassium fluorosilicate of 99.95% purity or higher. Also both
Monnier and Elwell have stated that equipment itself can lend to introduction of impurities into
the silicon [6, 19].
As mentioned in the Chapter 1, silicon can have several sources. Silica is abundant in its quartz
form with a purity of 99.5-99.8% with a price of 14c/kg [5]. However the difficulty of obtaining
the quartz, its purity and any processing done to purify it can vary the price, for instance the
quartz crystal grown by hydrothermal method for fused silica crucibles go at $1/kg [5]. It is
interesting to point out here that natural SiO2 tend to have lower B and P concentrations than
MG-Si [25] which is a huge advantage as these two elements are considered the most difficult to
remove. Fluorosilic acid H2SiF6 which is a byproduct of making P2O5 for fertiliser is produced at
20-40kg/ton of P2O5 which in turn is made at 40 million ton/year [19]. Fluorosilic acid is
neutralised to make K2SiF6 (lowest volatility) in a reaction with KF (in turn making dangerous
HF) or KOH (making H2O) and can be 98-99% pure at 6c/kg [5]. And finally a recent source for
silica comes from rice hulls, also a byproduct, at 20c/kg [5] where the impurities are
“comparable to mineral silica”, stating that it has a purity of 95-99%. Table 5 shows the
concentration of impurities for various sources and a comparison to what SoG-Si should be.
Please note Table 2 is not in agreement with the values given for SoG-Si in Table 1 and the
efficiency for SoG-Si in Table 2 was not stated in the reference [26]. Different references have
various ranges for specific elements but generally agree on the total impurity content.
Table 5 Impurity levels in ppm for different silicon sources and products [26]
Impurity Natural SiO2 Fluorosilic
acid Rice hulls MG-Si SoG-Si
B 0.3-0.5 1 2 5.4 <0.1-0.3
P 0.3-0.6 33 130 48 <0.1
Al 185 8 10 450 <0.1
Fe 56 13 20 3000 <0.1
Ti 210 - 3 130 <0.1
The choice of electrolyte, source of silicon and materials to construct the electrolysis cell can
determine the operating temperatures. This is an important parameter because this usually
dictates how much energy is consumed for the process of silicon extraction.
Page 29
20
2.4 Operating Temperatures
The methods to electrodeposit silicon can be categorised into three groups based on the operating
temperature.
1) Producing Si from a molten salt electrolyte at temperatures above melting point of Si
2) Molten salt electrolysis at a temperature range of 400-1150°C
3) Low temperature electrolysis at temperatures below 100°C
The melting temperature of silica is >1700°C so it is necessary to dissolve it in a low melting
temperature electrolyte for electrolysis. A binary or even ternary system is required. The authors
generally preferred fluorides of alkali and alkaline earth metals for their high decomposition
voltage, low melting points, high conductivity and low viscosity [6]. If the base electrolytes (in
cases where oxides were used) did not have these properties then fluoride additives were
introduced.
An attempt was made at gathering and summarizing the basic parameters in most studies made to
obtain Si or its alloys through electrolysis. Table 7 provided at the end of this chapter shows the
different methods separated into the three categories described above. The information thought
most useful were the electrolyte used; which will indicate how easy it is to obtain said
electrolyte, the operating temperatures; which will influence the materials used for the equipment
and energy consumption, the cathode material, the physical state of the deposited silicon (solid or
liquid) and the purity of the silicon; which will determine if further processing is needed.
Liquid deposition is more favourable to solid deposition since the rate of solid deposition would
have to be slow, on the order of 10’s of μm per hour for plane growth [19], to prevent protrusions
which trap inclusions. In liquid form the surface tension would exclude most of the solid
impurities and deposition rate can be increased by increasing the current density. However the
melting temperature of silicon is 1414°C [27] which means that to deposit in liquid form silicon
will have to be deposited at temperatures above 1414°C. When comparing to aluminium
deposition, the operating temperatures are between 940 and 1000°C [28]. Because of the
electrolyte composition and since the melting temperature of aluminium is 660°C [29] the
aluminium deposits as a liquid and can be siphoned off; an easy separation from the electrolyte.
Operating temperature depends on a couple of factors:
Page 30
21
The physical state of the deposited silicon: pure solid or liquid or alloy solid or liquid.
The melting temperature of the electrolyte mix.
To reduce operating temperatures, the silicon can be deposited as a liquid alloy by tailoring the
melting temperature of the alloy by changing its composition. For instance the silicon can be
deposited in a liquid copper-silicon cathode. As shown in the Cu-Si phase diagram in Appendix
A, the liquid alloy can start with approximately 5wt% silicon (lowest composition allowed while
still remaining liquid at 900°C) and can increase to about 20wt% silicon while still in the liquid
phase. In some cases alloy formation is done intentionally as the copper can also act as a getter
metal (i.e. solvent refining described in Section 2.2.3); the impurities that may have also co-
deposited with silicon preferably remain in the copper-rich phase rather than silicon during the
subsequent solidification. Two phases are created; a Cu3Si phase with dissolved impurities and a
pure phase of silicon. The alloy can then be crushed into particles the same size or smaller than
the grains of the pure silicon and can be separated from the rest of the alloy by flotation using
heavy media [13]. Which particular metal is preferred for alloying depends on the operating
temperatures, how well it works as a getter metal and in one case: if creation of shallow and deep
level recombination centres in the silicon can be avoided; the reason why tin is preferred to
copper or aluminium [4]. The ideal method would be to recycle the getter metal for further
silicon purification but the getter metal itself must be purified to prevent buildup of impurities
else it will not act as an effective getter metal. It has also been suggested that in cases where the
getter metal was Al, there is a distinct possibility that Al has substitution reactions with SiO2 to
make Si and Al2O3 and electrolysis does not play a part [25]. In this case there is a loss of getter
metal although a higher than expected silicon recovery occurs. However, if the silicon is to be
deposited in a solid state then the operating temperature is determined by the electrolyte
composition.
Regardless of Si being deposited as liquid or solid, lowering the operating temperatures is
beneficial from several aspects:
Reduced vapourisation losses of electrolyte
Lower energy consumption
Easier operation and wider choice of construction materials
The electrolyte must remain liquid at the operating temperature and if necessary, its composition
be adjusted for this purpose. For instance cryolite melts at 1012°C±2 [30] but when 10wt%
Page 31
22
alumina is dissolved in it the melting temperature drops to ~980°C, eutectic being at 20wt%
alumina at 960°C [31]. Temperature can be further reduced by adding aluminium fluoride AlF3.
Additives like aluminium fluoride and alkaline and alkaline earth metal halides are also added to
decrease viscosity and increase conductivity. When silica is added to cryolite the melting
temperature does not decrease significantly; preliminary experiments showed the eutectic to be
around 998°C, so additives should be added to reduce operating temperatures. According to the
cryolite-lithium fluoride phase diagram [32] shown in Appendix A, 13.8wt% of LiF dissolved in
cryolite can reduce the melting temperature of the whole composition to 900°C; one reason why
it was chosen as an additive for this thesis. Table 6 below shows the effect of a number of alkali
and alkaline earth fluorides on melt properties when electrolysing alumina in cryolite to form
liquid aluminium [33]. What must be determined is if the same table is applicable to cryolite
melts with silica; although silicon is next to aluminium on the periodic table, silicon oxide has a
slightly more acidic nature than aluminium oxide, silicon ions needs 4 electrons for reduction to
metal compared to the 3 for aluminium and silica reactions with cryolite leads to more phases
being formed than when alumina is added to cryolite.
Table 6 Additive effect on cryolite-alumina electrolyte [33]
Additive Conductivity Density Viscosity Liquidus
Temperature
Al Metal
Solubility
Surface
Tension
Vapour
Pressure
CaF2 ↓ ↑ ↑ ↓ ↓ ↑ ↓
AlF3 ↓ ↓ ↓ ↓ ↓ ↓ ↑
MgF2 ↓ ↑ ↑ ↓ ↓ ↑ ↓
NaCl ↑ ↓ ↓ ↓ ↓ ↓ ?
LiF ↑ ↓ ↓ ↓ ↓ ↑ ↓
Ideal
Additive ↑ ↓ ↓ ↓ ↓ ↓ ↓
Temperature
increase ↑ ↓ ↓ --- ↑ ↓ ↑
Additives should increase conductivity of the electrolyte to reduce energy consumption, decrease
density to allow floatation on top of deposit, decrease viscosity to increase mobility of the ions
that are to deposit, decrease liquidus temperature to decrease operating temperature and hence
Page 32
23
energy consumption, decrease deposit solubility in melt to decrease loss of yield, decrease
surface tension to improve wetting of the melt with the electrodes and reducing polarisation
resistance, and decrease vapour pressure to avoid loss of electrolyte. As can be seen LiF meets
all but one of the criterion and hence an ideal choice as an additive.
It is also important to mention that all the additives listed in the table decreases the solubility
limit of alumina in cryolite. This can have a negative consequence in that the anode effect will
happen sooner if the alumina concentration in the melt is not monitored. A study done by Zhang
et al. [34] looked at the possible complexes formed by alumina to explain how the solubility
changes with cryolite ratio and with additives. Cryolite ratio is defined as the mole ratio
NaF/AlF3; 3 is considered a neutral melt, ratio below 3 is acidic and above is basic. This affects
the complexes formed like so:
In acidic melt: 1/3 Al2O3 (s) + 2 NaF (l) +
4/3 AlF3 (s) = Na2Al2OF6 (l)
In neutral melt: 2/3 Al2O3 (s) + 2 NaF (l) +
2/3 AlF3 (s) = Na2Al2O2F4 (l)
In basic melt: 1/3 Al2O3 (s) + 2 NaF (l) +
1/3 AlF3 (s) =
1/2 Na4Al2O2F6 (l)
As can be seen from the reactions above, when the ratio between NaF:AlF3 increases from acidic
to neutral the solubility of alumina increases, but when the melt becomes more basic the
solubility decreases, illustrated in Figure 7 showing experimental results.
Figure 7 Alumina solubility versus cryolite ratio
0.04
0.05
0.06
0.07
0.08
0.09
0 2 4 6 8 10 12 14
Dis
solv
ed A
l 2O
3 (
mola
r fr
act
ion
s)
Cryolite Ratio (NaF/AlF3)
Page 33
24
The points in red indicate cryolite ratio at 3 and near 8 (explained in Chapter 4). The same
authors [34] found that LiF is a basic additive so it can be expected to decrease the solubility
limit of alumina in cryolite, as is observed in experiment [28].
2.5 Choice of Electrolyte
The Hall-Héroult process has been using cryolite as the main electrolyte in the electrolysis of
alumina to obtain aluminium. Numerous studies have been done on the cryolite; its physical
properties, chemical properties and handling of the material. Cryolite itself is described as an
“uncommon mineral” and the largest deposit in Ivigtût, Greenland has been depleted. However it
can be manufactured through various routes, three of which are shown below [35, 36]:
12HF + Al2O3.6NaOH 2(AlF3.3NaF) + 9H2O
2AlF3.3HF + 6NaOH 2(AlF3.3NaF) + 6H2O
AlF3 + 3NH4F + 3NaCl Na3AlF6 + 3NH4Cl
Cryolite and its additives are known to produce toxic HF gas but the infrastructure to treat the
gas has already been put in place at aluminium industries. With so much physical and chemical
information about the electrolyte already available, groundwork for the use of this chemical for
electrolysis is already laid out, and with the ease of manufacture of the electrolyte, one can
conclude it would be convenient and worthwhile to investigate the viability of using cryolite as a
base electrolyte in the electrolysis of silica to obtain silicon. Electrolysis work has already been
done using cryolite and solid silicon with upto 3-5N purity has been obtained (Table 7) therefore
cryolite has been deemed a practical electrolyte worth further investigation [6, 37, 38].
The additive chosen for this thesis is lithium fluoride and a composition, based on the Na3AlF6-
LiF phase diagram (Appendix A), was chosen to reduce the melting temperature of the CR+LiF
mixture to 900°C. Firstly, the melting temperature of pure cryolite has been reported as 1012°C ±
2 [30] depending on the impurity concentration of the cryolite used in the various studies.
Secondly the Si-Cu phase diagram (found in Appendix A) shows that the minimum temperature
at which Si-Cu alloy will remain liquid between 10-30 mol% of Si at is 860°C. Therefore a
working temperature between 900-1000°C keeps the alloy liquid and the reaction kinetics
relatively fast, while not having to work at overly high temperatures thus saving energy. Hence a
Page 34
25
melt composition of 13.8wt% LiF with 86.2wt% cryolite was used as the basis for the
electrolyte.
A special mention should be given to Ming et al. [39] who have already deposited silicon in the
solid state from a cryolite and lithium fluoride melt at a purity of 99.5 to 99.97% thereby
demonstrating that the electrolyte used in this study is functional. However fundamental
properties of the CR+LiF and silica melt cannot be found in literature and this thesis can
contribute knowledge about three properties for this melt and the way they have been measured.
2.6 Measuring Properties of Electrolyte
Pertaining to electrolyte production of metals, the following properties of the system are of
interest:
Density
Electrical conductivity
Electronic and ionic conductivities
Viscosity
Solubility and diffusivity of species being deposited in electrolyte.
In the present study, density of CR-LiF-SiO2 melts together with dissolution behaviour of SiO2 in
CR+LiF melt is investigated experimentally.
2.6.1 Determining Dissolution Behaviour
Operating parameters have an effect on the macrostructure. The goal of the work done by Elwell
et al. [19, 21] was to get coherent deposition of layers of silicon on the cathode. For low
concentration of silicon ions and high current density they got powdery or dendritic growth. For
low concentrations, the silicon ions become depleted at the cathode surface and by the time new
ions diffuse through to the cathode they start a new grain leading to powdery deposition or even
deposition of unwanted cations (although this can be avoided using pulsed electrolysis to allow
for transport of silicon ions to cathode surface). For high current density the ions are quickly
placed in the easiest to grow crystal direction which leads to dendritic growth. Increasing the
concentration of the silicon ions for a higher current density or decreasing the current density for
a lower silicon ion concentration gave the researchers coherent layers on a silver cathode.
Page 35
26
However the concentration of silicon ions had to be much higher to get coherent deposition on
graphite cathodes. The current density determines how fast layers of silicon can grow and in the
work done by Elwell they managed to grow a 3mm thick silicon layer on an area of
approximately 2cm by 1cm silver plate in 4 days. Because the deposition took so long, the
vapours from the electrolyte reacted with the Inconel tube the graphite crucible was set up in. Fe
and Cr vapour compounds dissolved in electrolyte and were co-deposited with silicon
compromising the purity of the deposit.
The importance of the solubility limit is highlighted above; the researchers could only grow a
coherent deposit after a certain concentration of silicon ions in the melt was met. Also the anode
effect must be avoided; when the ions to be deposited at the cathode are depleted, an insulating
gas layer on the anode is formed and the cell voltage increases and the current flow decreases. In
the aluminium electrolysis cell the gases are fluorocarbons (in addition to COx) and have to be
dealt with [28]. Knowing solubility limits help avoid this; anode effects were a common
occurrence at the beginning of the era of electrolysis of alumina but with saturation limits used to
advantage a cell can run a whole day without an anode effect. It is interesting to note that
Grjotheim et al. discovered that the solubility of silica in cryolite increases significantly with the
addition of alumina with a maximum at 69wt% SiO2 for 14wt% Al2O3 [40]. Of course this
increases the probability of aluminium co-depositing with silicon. But also crucial to avoid is
oversaturation; too much silica could lead to silica particles sinking in the melt form inclusions in
the liquid deposit.
The importance of the property of diffusivity or mass transfer coefficient is also stressed; the
higher the diffusivity of an ion the less likely it is to be depleted at the cathode and the more
likely a coherent layer will grow. The k values can also be used to determine the limiting current
that can be used to avoid the anode effect. A secondary significance of mass transfer was shown
in a couple of studies [41, 42] where the dissolution rate of alumina particles into the cryolite
melt determined whether the particles would completely dissolve before it reached the bottom of
the crucible or sink into the molten aluminium at the bottom of the crucible and become an
inclusion or be lost in the sludge.
Diffusivity, D, and mass transfer coefficient, k, are related as [43] with a constant of
proportionality called the effective film thickness, δ, which needs to be calculated based on k and
Page 36
27
D values. Depending on the difficulty of handling the material and equipment at hand, one
property may be favoured over the other for measurement. Both properties are related to the
dissolution rate of a solute into a solvent. With studies of alumina in cryolite several researchers
found that some additives such as LiF, CaF2, MgF2, AlF3 and NaCl slowed the dissolution rate
while others like NaF increased it [28]. It has been found that silica dissolution into cryolite was
very slow indeed [44] but no quantitative measurements were made. This will be one of the areas
this thesis will be exploring along with solubility limits.
2.6.1.1 Measuring solubility
Although Grjotheim et al. mention that the maximum solubility of silica in cryolite is 5wt% at
1000°C [45] they have not mentioned what experimental procedure they had used. For alumina
solubility in cryolite Zhang et al. [46] added excess alumina to cryolite and the mixture was
stirred with a platinum rod at regular intervals of 20 minutes. A small sample was then quenched
onto an alumina rod which was then digested (although the solvent they used is not mentioned)
and analysed using inductively coupled plasma (ICP). The total aluminium ion content was
compared to a “blank” (a pure cryolite sample) and the additional alumina in the sample was
calculated which gave an indication of the solubility. It is important to note here that AlCl3 is a
well-known solvent for dissolving cryolite and it does not dissolve alumina [47]. The problem
with applying this method to silica is that when silica dissolves in cryolite it reacts to form a
number of compound that do not dissolve in AlCl3 solution. When using AlCl3 solution a gel like
substance was created which was believed to be silica gel. This does not give a true
representation of the solubility of silica in cryolite.
Jentoftsen et al. [48-50] found the solubility of FeO, FeAl2O4, NiO and NiAl2O4 in cryolite by: 1)
dissolving an excess amount of each oxide in cryolite and then quenching the sample, 2)
carbothermically reducing all excess oxides, 3) digesting the sample with hydrochloric acid to
dissolve Fe or Ni metal and the using ICP to determine Fe or Ni content in the sample. This is
not feasible for silica in cryolite because after dissolving in cryolite silica starts reacting with it.
Then they found the solubility of TiO2 in cryolite by: 1) dissolving an excess amount of each
oxide in cryolite and then waiting 90 minutes for the TiO2 powder particles to settle in the melt
and then quenching the sample, 2) using X-ray fluorescence (XRF) which does an elemental
analysis to find titanium content. Solubility of TiO2 in cryolite at 1020°C is 5.2wt%. This method
Page 37
28
is clearly not very accurate as settling of very fine particles can be very slow. In particular, in the
cryolite-silica system, the difference in density is small, making the settling even slower. Finally
Jentoftsen et al. found the solubility of CuO and Cu2O in cryolite by: 1) dissolving an excess
amount of each oxide in cryolite and then quenching the sample, 2) digesting the sample and
analysed using ICP. This method is again not applicable for silica.
2.6.1.2 Measuring diffusivity/mass transfer coefficient
The importance of the need to know diffusivity and/or mass transfer coefficient (D or k) has
already been emphasised above. There is an extensive review on different methods of measuring
diffusivity [51] and a few of them were tried to see if they were applicable to cryolite melts. One
of the main problems with cryolite is that it is very corrosive and so far only graphite has been
found not to react with it. Many researchers have mentioned using platinum crucibles to contain
it (for very small samples) and rods to stir it.
Capillary reservoir methods involve having a reservoir which contains the melt and the ion
whose diffusivity needs to be measured, and a capillary tube with only the melt [51]. The
capillary is left in the reservoir and the ion in question diffuses into the tube because of a
concentration gradient. The tube is subsequently quenched and the concentration of the ion is
measured. Some authors [52, 53] used a radioactive isotope to track the movement of an ion
through the melt by gamma-ray attenuation, however there is no silicon isotope with gamma
decay. Chronopotentiometry is another method used to measure diffusivity but because silicon
ion mobility is much slower than sodium and fluorine especially since it creates complexes in the
melt it does not contribute significantly to the measurements therefore is hidden in the variation
of the results. A final method described by Kubíček [51] was forced convection dissolution of
rotating discs and measuring the loss of weight with time as shown in Figure 8. Knowledge of
viscosity and angular velocity will allow determination of diffusion coefficient. This method was
modified to a stationary rod with natural convection and used for this thesis, Figure 9. The mass
loss with time can be fitted to the general mass transfer equation to determine k.
m = mass of silica rod dissolving in melt (g) (only considering the portion inserted in the melt)
t = time of dissolution (h)
Page 38
29
k = mass transfer coefficient (cm/h)
A = area of silica rod exposed to the melt (cm2)
CS = saturation concentration of silica in melt at the temperature of experiment (g/cm3)
CB = concentration of silica in the bulk at time t (g/cm3)
Figure 8 Schematic for measuring diffusion coefficients by rotating disc method [51]
Figure 9 Modified schematic to observe dissolution behaviour for this study
Page 39
30
2.6.2 Measuring density
Once the silicon is deposited, it should be protected against the atmosphere to prevent its
oxidation. When aluminium is deposited as liquid, it sinks to the bottom and the floating
electrolyte prevents it from oxidising. It is necessary that either the liquid silicon (or alloy of) has
a higher density than the electrolyte or that the surrounding atmosphere is inert. Work done by
Grjotheim et al. [45] is shown in Figure 10; the density variation with alumina and silica
additions to pure cryolite. For both instances increase in the oxide concentration, the researchers
measured a decrease in density.
As mentioned, Grjotheim et al. [45] determined the solubility of silica in cryolite at 1000°C to be
5wt% which is why the additions for silica stop at 6wt%. Both oxides decrease density but silica
slightly less than alumina, as shown in Figure 10. The decrease in density for alumina-cryolite
melts is explained by the aluminium complexes (with oxygen and fluorine) that form in the melt
which has a large radius [33]. However for silica addition to cryolite Sokhanvaran et al. [54]
found an increase in density.
Figure 10 Density of the electrolyte with alumina and silica additions to cryolite [45]
2.03
2.04
2.05
2.06
2.07
2.08
2.09
2.10
2.11
0 2 4 6 8 10 12
Den
sity
(g/c
m3)
Oxide Concentration (wt%)
Silica
Alumina
Page 40
31
For comparison purposes the density of liquid aluminium in the rage of 933 – 1190K (660 –
917°C) is:
where Tref is the melting temperature of aluminium at 933.47K [29]. This indicates that the
molten aluminium will be denser than the electrolyte. The reason why aluminium is considered
here is because of all the considered alloy metals, aluminium has the lowest density so if the melt
has lower density than liquid aluminium then the other liquid metals like copper and tin will not
cause a problem. The bigger the difference between the densities the better because the less
likely the silicon liquid globules that may form due to disturbance of the bath remain suspended.
One author suggested that if the liquid silicon or silicon alloy is less dense then the cell can be
kept in an inert environment or use Down’s cell similar to the one used for sodium electrolysis
[4]. Another way to make sure that oxidation is at a minimum is to deposit the silicon as a solid
(or alloy of) at the cathode which can be positioned so that the electrolyte is floating on top. It
should be noted that in most instances in Table 7 given below solid alloying of silicon was
unintentional; the silicon reacted with the cathode to form an intermetallic silicide. In some cases
[15, 55] the silicide formation was used advantageously to create a hard coating e.g. iron silicide.
The most common way of measuring density is using Archimedean principle. Grjotheim et al.
[45] measured the density of Na3AlF6-SiO2-Al2O3 using the buoyancy method with Pt-10Rh
sphere. This method was modified for this thesis by using a nickel ball for the sink instead of
platinum; nickel reaction with the CR+LiF melt is negligible.
Page 41
32
Table 7 Conditions of various studies on electrolysis of silicon
Year Authors Electrolyte Silica Source Cathode Silicon form Working
Temperature
Silicon
Purity
Molten Salt Electrolysis 400-1150°C
Pure Silicon (or attempt)
1854 [56] Deville AlCl+NaCl
Accidental silicon
impurity in the
melt
Graphite Crystalline silicon ? 10.3%
1854 [56] Deville NaF + KF + SiO2 SiO2 Platinum Platinum silicide (intermetallic;
melts and disappears into the bath) ? -
1865 [57] Ullik KF+K2SiF6 K2SiF6 - Metallic K and Si (immiscible
alloy?) ? -
1884 [58] Gore K2CO3+K2SiF6 K2SiF6 Platinum Platinum silicide ? -
1888 [59] Hampe NaSiO3 NaSiO3 Carbon
Amorphous silicon (extracted by
reacting with hydrochloric acid and water)
? ?
1934,39 [60, 61]
Dodero
Alkali and alkaline earth metal
silicates with fluoride additives to
lower temp and viscosity.
From electrolyte Iron
Cathode reacts with deposited Si.
Some alkali metal co deposited as
well.
800-1250°C 72%
1951 [62] Wartenberg et al. NaCl + K2SiF6 K2SiF6 Nickel Brown needles with 0.1%Fe,
0.05%Ni, 0.1% Na & K 900°C >99%
1959 [63] Stern at al. Alkali chloride (or mixture) +
K2SiF6 (electrorefining) SiC anode Steel
Solid crystalline (easily broken
away from cathode). 400-1000°C 92-99.3%
1959 [64] Olstowski Alkali and/or alkaline earth fluoride
or AlF3 mixtures + Na2SiF6 Na2SiF6 Iron or copper
Initially intermetallic silicide and
after saturation some silicon 450-800°C ?
1957-64 Monnier et al. Na3AlF6 + SiO2 (electrorefining) SiO2 MG-Si at anode to deposit Si
at cathode Solid silicon 1000°C 99.99%
1968 [65] Jorgensen
Thin beds of solid SiO2 (obtained
from oxidation of p-type silicon wafers – this experiment was done
to obtain reduction voltages.)
SiO2 Silicon Polycrystalline silicon 800-880°C Depends on
purity of
silicon wafer
1971 [55] Lyakhovich et al. Na2O + SiO2 SiO2 Armco iron or steel Intermetallic of iron and silicon 1050-1150°C -
1975-77
[66-68] Cohen Alkaline fluorides + K2SiF6
Silicon cathode
5N purity Monocrystalline Silicon Epitaxial monocrystalline silicon 600-850°C ?
1980-81 [19,
21]
Rao, Elwell &
Feigelson
Flinak-Eutectic LiF+NaF+KF and
K2SiF6 or
LiF+KF and K2SiF6
K2SiF6 Silver or graphite Polycrystalline silicon ~745°C 99.98-
99.999%
1981 [19,
69]
Rauh and separately
De Mattei High temperature fluorides K2SiF6 Silicon Crystalline silicon 700°C ?
1981 [20] Elwell NaF + CaF2 + SiO2 SiO2 Graphite Discrete silicon particles near
cathode. 1150°C ?
Page 42
33
Table 7 (Cont’d) Conditions of various studies on electrolysis of silicon
Year Authors Electrolyte Silica Source Cathode Silicon form Working
Temperature
Silicon
Purity
1981, 83
[70, 71] Olson & Carleton LiF + KF + K2SiF6 (electrorefining) MG-Si-Cu alloy
anode Si with Cu addition at anode Solid polycrystalline silicon 750°C 99.999%
1981 [72] Olson & Kibbler Na3AlF6 + LiF + AlF3 + BaF2 SiO2 Liquid Sn Si phase separate in Sn with Al, Ba
and Sn impurities. ? ?
1996
[73] Stubergh and Liu
Bytownite (mixture of CaAl2Si2O8
and NaAlSi3O8) + Na3AlF6 Bytownite Graphite
Solid polycrystalline silicon with Fe and Si3Fe impurities. Fe from
bytownite carbothermically reduced
by graphite.
970°C 99.79-
99.98%
2005
[74] Yasuda et al. LiCl + KCl + CaCl2 SiO2 Silicon
Solid amorphous and microcrystalline silicon; sponge-
like particles at 50nm
500°C ??
Alloy of silicon
1891 [75] Minet NaCl+Na3AlF6+(Al2O3 or FexOy) +
SiO2 SiO2 Carbon
Al-Si alloy (liquid?) or solid Fe-Si
alloy 700-1000°C -
1893 [76] Warren Alcohol solution + SiF4 SiF4 Mercury Silicon amalgam ? -
1933 [77] Batashev & Zhurdin Na3AlF6 + Al2O3 + (Al2Si2O5(OH)4
or SiO2)
Al2Si2O5(OH)4 or
SiO2 Al-Si alloy ?
27% of silicon in
alloy
1957-64 [18,
23, 24, 78-
81]
Monnier et al.
Mixture of NaF, LiF or KF and
K2SiF6 (electrowinning). Other
electrolytes were tried; please see paper for full detail.
K2SiF6. Other
sources were used
like SiO2 in
different
experiments
Liquid Cu Si-Cu alloy 1000°C 4-25%
silicon
1970-75 [45,
82-89]
Grjotheim, Matiasovsky, Fellner
& Bøe (et al.)
Mainly Na3AlF6 + Al2O3 + SiO2; properties of and deposition of
alloys of Silicon
SiO2 Various: e.g. graphite, copper, aluminium and
nickel
Liquid alloys of silicon. Solid
silicon deposited on graphite. >1000°C -
Low temperature electrolysis below 100°C
1966 [90] Zyazev & Ezrielev propylene
glycol or pyridine + KI, NH4Cl SiF4 or SiCl4 Platinum Amorphous silicon ? ?
1976 [15] Austin
Aprotic dipolar organic solvent one
of: propylene carbonate, tetrahydrofuran, acetonitrile,
dimethyl sulfite,
tetramethylenesulfone, ethylenecarbonate, N,N-
dimethylacetamide, and
dimethylformamide
SiX4 (X = Br, Cl
or I) or HnSiX4-n For electroplating purposes
Amorphous Si:H (Other electrolytic works in organic
mixtures have been done but all
obtain a-Si:H which cannot be used for high power solar cells because
of lower efficiency.)
20-100°C -
Melting temperature above Silicon >1400°C
1981 [22] De Mattei, Elwell &
Feigelson BaO + SiO2 + BaF2 SiO2 (99.5% pure) Graphite
Liquid silicon (NB: efficiency was
only ~40%) >1450°C 99.97%
Page 43
34
Chapter 3 Experimental
This chapter explains the experimental procedures, apparatus and materials used to obtain three
properties with two types of experiments:
Dissolution behaviour:
Solubility limit of silica in the CR+LiF electrolyte at temperatures between 900 and
1000°C.
Dissolution rate of silica in the CR+LiF electrolyte at temperatures between 900 and
1000°C.
Density measurements:
Density of the CR+LiF electrolyte with varying silica concentrations at temperatures
between 900 and 1000°C, using Archimedean principle.
Each experimental technique will be presented in two main sections. For each property
measurement, the most adaptable method was employed based on the salt that was being used
and the equipment at hand. For instance the corrosive nature of the cryolite melt and the working
temperature had to be taken into account when designing experimental methods. Under the
dissolution behaviour section various approaches that were examined in the present study will
also be discussed to highlight the difficulty of working with cryolite.
A further section is added to explain an attempt at measuring conductivity of the CR+LiF melt
with varying silica concentration. The Van der Pauw method was unsuccessfully employed and
what transpired and why it did has been described in this section.
Page 44
35
3.1 Dissolution behaviour
The experiment explained in this section was used to quantify the dissolution behaviour and
solubility limit of silica into the CR+LiF electrolyte at various temperatures.
3.1.1 Examined Approaches
1. SEM and EDS of quenched CR+LiF+Silica samples
An initial study was carried out to determine solubility limit of silica in the CR+LiF sample. The
concept was to dissolve different amounts of silica in the CR+LiF melt and quench the crucible
at various temperature, then observe under the SEM the different phases that would be formed if
the sample was oversaturated with silica. EDS with point analysis could be used to determine
concentration of silica in each of the phases including the saturated liquid. However even though
the sample size was varied from 2 to 15 grams, water quenching was not quick enough as the
subsequent analysis revealed. The crucible used was thick graphite; to contain the corrosive
cryolite and prevent being completely consumed by oxygen. SEM and EPMA analysis showed
that the silicon is concentrated into globular phases within a matrix as shown in Figure 11. The
SEM image shows a sample containing 6wt% silica quenched at 1000°C and on the right
elemental mapping of silicon. EPMA analysis of the sample showed the globular phases within a
matrix had high concentrations of silica, 20-40wt%, whereas the matrix itself had 1-2wt% silica
content. Previous authors had concluded that cryolite, when oversaturated, will react with the
silica to form combinations of sodium, aluminium and silicon oxides, (Na.Al.Si)Ox, called albite,
jadeite and nepheline [91] which are high in silica content. Which particular mineral is formed is
time dependent and it was concluded that the cooling rate was too slow allowing the formation of
the albite/jadeite/nepheline phases.
Visual observations of the sample showed a “shiny” phase which could be observed at 4-6wt%
silica depending on the sample size as shown in Figure 12. This value depends on the size of the
sample, therefore the smaller the sample the more accurate the method. However it can be only
estimated to the nearest whole number and this may lead to the same value for a range of
temperatures, therefore this method was deemed too imprecise to predict solubility limit.
Page 45
36
a) b)
Figure 11 a) SEM Image of a 6wt% sample b) Si EDS of sample
Figure 12 Sample size of 2g containing 6 wt% silica
“Shiny” phase believed to
be high in silica content
Page 46
37
2. ICP Analysis of oversaturated CR+LiF+Silica samples
Another attempt at measuring solubility limit was made using a chemical approach; 10wt% of
silica was added to CR+LiF melt to make 10g samples. Blank CR+LiF samples with no silica
were also made for comparison. These were quenched at different temperatures, crushed to
powder and leached with 15wt% AlCl3 solution which can dissolve cryolite. The rationale was
that any silica dissolved in the cryolite will be able to stay in solution whereas the undissolved
silica will remain as a solid and can be filtered out. However the silica reacted with cryolite to
form oxide compounds of albite/jadeite/nepheline phases which reacted with the AlCl3 solution
to form silica gel and blocked the filter paper corrupting the results.
The mass of the residue from filtering out the powder and solution indicated a decrease in
solubility with increasing temperature which is counterintuitive. Furthermore ICP of the filtrate
showed very little silica content and when a mass balance was conducted, the masses of the
residue and that calculated from the filtrate did not add up to the initial mass of silica in the
sample. The main reason for this is the loss of silica content as silica gel. Moreover the excess
silica is not believed to remain unreacted in the CR+LiF melt; it is supposed that it will further
react with cryolite to form the aforementioned albite/jadeite/nepheline phases. Therefore a true
solubility limit cannot be determined and this method was deemed impracticable.
3. Diffusivity profile in silica tubes
A small silica tube was inserted in a molten CR+LiF mixture to draw molten salt up the tube due
to capillary action and slowly dissolve into the cryolite. Ideally, a diffusion profile along with a
saturation concentration can be obtained if the CR+LiF inside the tube is analysed with
EDS/EPMA, Figure 13. In that case, the method could give both diffusion coefficients and
solubility limits. Different tube inner diameters and different immersion times were
experimented to obtain a reasonable profile.
Page 47
38
Figure 13 Anticipated diffusion profile across a silica tube with CR+LiF melt
The whole crucible had to be quenched since extracting only the silica tube would disturb the
profile as some of the liquid would flow out the bottom. Unsuccessful attempts were made to
seal the end of the silica tube but quenching the whole crucible was deemed safer and more
accurate. The graphite crucible had outer diameter of 6cm, inner diameter near the top of the
crucible of 5cm (inside crucible is conical in shape), height of 8cm. Again due to slow cooling
there were phase separation and EPMA point analysis did not produce any meaningful results.
However with EDS average area analysis, Figure 14, a reasonable profile of silica concentration
across the tube was obtained, Figure 15. The black squares in Figure 14 indicate the areas that
were analysed for silicon (which was then converted to silica concentration) and the centre of the
box was taken as the distance from the edge. Area analysis ensured that the phase separation due
to albite/jadeite/nepheline formation was averaged out and Figure 15 shows a reasonable value
for saturation concentration of 5-7wt% (as opposed to 20-40wt% in the mineral oxide formed).
However as can be seen from the graph it is not a parabolic curve as was expected.
8cm
6cm
Page 48
39
Figure 14 Diffusion Experiment Sample
Figure 15 SiO2 profile across a 2mm tube held in CR+LiF melt for 30 minutes
The silica walls dissolve away creating circular flow motion of the melt near the walls which
disturbs the diffusion profile as shown in Figure 16. There is a significant problem involved in
this experimental method; natural convection due to ΔT (temperature) and Δρ (density). ΔT can
be eliminated if the sample is small enough and placed near the centre of the furnace. Δρ ,
0
1
2
3
4
5
6
7
0 500 1000 1500 2000
SiO
2 (
wt%
)
Distance (μm)
Page 49
40
created due to silica dissolving into the CR+LiF melt and making a denser composition, can be
eliminated by using the capillary reservoir methods.
Figure 16 Effect of convection flow on the concentration profile
4. Capillary reservoir methods
This approach involves having a reservoir and a capillary tube made of graphite, shown in Figure
17. The reservoir contains a CR+LiF and silica melt, whereas the capillary tube is filled with
CR+LiF only. The idea is that the silica from the reservoir will diffuse into the capillary tube; the
highest concentration at the opening of the tube and a concentration profile into the tube. This set
up allows for the higher density silica+electrolyte melt to be near the bottom.
Figure 17 Capillary reservoir to measure diffusivity
Capillary tube:
melt (no silica)
Reservoir: melt
with silica
Page 50
41
However this method was attempted without success; one of the main problems was that graphite
floats on cryolite so the capillary tube cannot be dropped into the melt and expect diffusion to
happen. There are other drawbacks to this method:
1. Convection due to temperature and density gradient. These can be overcome if the setup
is small enough and so it is all at one temperature and if the higher density solution is
created at the bottom of the capillary tube so as to prevent fluid motion.
2. Boundary conditions are uncertain even though at the entrance [CSi] is defined as the Si
concentrations in the reservoir.
3. Wall effects that create a drag on the diffusion.
4. During solidification the volume contraction can change the distribution affecting results.
3.1.2 Selected Method
Quartz rods were immersed in the electrolyte and the mass loss and change in radius were
measured against the immersion time. The results were treated to yield solubility limit and mass
transfer coefficient of silica ions. Cryolite is a very corrosive compound, thus graphite crucible
was used. However the graphite at elevated temperatures will react with the oxygen inside the
furnace so argon gas purging was necessary to keep the atmosphere inside the furnace as inert as
possible. Quartz rods (impurity less than 25ppm) 2.5cm in diameter and 10 cm in height were
inserted in the molten sample size of 200g CR+LiF through a hole in the crucible cap (also of
graphite). The schematic in Figure 18 shows the overall setup used to observe dissolution
behaviour. The quartz rod was left resting at the bottom of the crucible for various periods of
time. The fume hood is necessary for this set up since silica reacts with the CR+LiF melt
producing SiF4 gas.
It should be noted that a study was done by Gairola et al. [92] on how the dissolution behaviour
differs when a nickel cylinder is suspended in liquid aluminium versus resting it at the bottom of
the crucible. Resting the cylinder at the bottom of the crucible interrupts the free convection flow
leading to different dissolution rates; the suspended cylinder dissolves quicker. In this study it is
believed that by resting the cylinder at the bottom of the crucible accelerated dissolution around
the edges of the cylinder is eliminated hence the dissolution is more even around the cylinder.
Page 51
42
Mass loss and reduction in radius of the rods were measured after immersion. Mass
measurements are preferred over radius measurements because as Figure 20 shows 1) radius can
vary along the length of the immersed portion of the rod; the biggest change in radius is near the
surface of the melt where the velocity of the melt is highest and 2) the immersed portion did not
have a smooth surface. One experiment consisted of keeping the temperature constant and was
run for a total of 8 hours, with different dissolution time periods for 6 rods: 15min, 15min,
30min, 1h, 2h, 4h. Figure 19 shows the rods after immersion; from left to right the cumulative
immersion time increases. After removing from the crucible, the rods were cleaned with AlCl3
solution by immersing them into the solution and using a hot plate and magnetic stirrer for 8
hours to dissolve any solidified melt on the surface of the rods. This procedure was done for five
temperatures: 920, 940, 960, 980 and 1000°C and repeated three times.
Argon Gas
Inlet
Furnace
Silica
Rod
Electrolyte
Figure 18 Schematic of the experimental setup for dissolution behaviour study
Page 52
43
Figure 19 Rods after the experiment
Figure 20 Immersed portion of rod
0.25 h 0.5 h 1 h 2 h 4 h 8 h
Page 53
44
For each temperature, the cumulative mass loss was plotted against time and the curve fitted to
the mass transfer equation described below to find the values for mass transfer coefficient and
solubility limit.
To calculate the mass transfer coefficient k and solubility limit CS from mass loss-time data, the
general mass transfer expression, Equation (6), was integrated to get an expression where mass
of the silica rod, m, was expressed in terms of time, t.
General mass transfer equation:
(6)
where:
m = mass of silica rod dissolving in melt (g) (only considering the portion inserted in the melt)
t = time of dissolution (h)
k = mass transfer coefficient (cm/h)
A = area of silica rod exposed to the melt (cm2)
CS = saturation concentration of silica in melt at the temperature of experiment (g/cm3)
CB = concentration of silica in the bulk at time t (g/cm3)
The area A and bulk concentration CB were expressed in terms of m and the equation was
integrated over time (full derivation is found in Appendix B) to yield Equation (7).
(7)
Where C1, C2 and C3 are constants defined as:
|
√ √
√ √ |
√
( √
)
V = volume of melt is assumed to be constant with dissolution of silica because the maximum
amount of silica dissolved is less than 9g into a melt of 200g. Also at each time (15min, 30min,
1, 2, 4, 8hrs) a fresh rod of initial radius of 1.25cm is inserted and the change of radius is less
than 0.06cm.
Page 54
45
h = height of melt after the rod is inserted into melt, also assumed constant for reasons given
above (associated change in height due to change in volume is less than 0.3cm).
Others constants: m0, ρ, π, CS, and k.
m0 = initial mass of silica rod inserted in the melt
ρ = density of silica rod
In order to fit the data into Equation (7), the MS Excel Solver program was used. CS and k were
set to vary, while the sum of square of errors, defined as the difference between calculated and
measured masses, was minimised. It should be emphasized that the cumulative mass loss at 8
hours was not used as the saturation limit (CS) because it was assumed at 4 hours it had already
reached saturation and further dissolution was in fact reaction of silica with the cryolite. The k
and CS were also calculated using data from one experiment that ran for 16 hours and then using
data points up to 4 hours to see how this affected the calculated k and CS values.
In order to test the reproducibility of the results, dissolution of quartz in CR+LiF melt was
repeated three times at 1000°C. At 1000°C one can expect the highest velocity of fluid causing
the most amount of uneven dissolution. Figure 21 showed initially that the experiment was
indeed reproducible.
Figure 21 Reproducibility of dissolution behaviour experiment
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8
Cu
mu
lati
ve
Mass
Loss
(g)
Time (h)
exp1
exp2
exp3
Page 55
46
During the experiment, SiF4 can be released from the reaction shown in Equation (8) [91]:
SiO2 + Na3AlF6 = xNa2O.yAl2O3.zSiO2 + NaF + SiF4↑
(8)
The values for x, y and z indicate which of albite, jadeite or nepheline is formed. Since it is
thought that jadeite is formed initially and with time decomposes to albite and nepheline, all
three may be present in the melt.
The SiF4 is thought to further react with water vapour as shown in Equation (9):
SiF4 + H2O SiO2 + HF
(9)
which tends to shift to the right at temperatures above 800°C (HSC Chemistry 6.1) when the
Gibbs’ free energy becomes negative. The SiO2 deposits as a white solid that was observed on
the top part of the silica rod. The deposit could not be dissolved away with AlCl3 solution so it
cannot be cryolite. XRD analysis (Appendix C) on the deposit showed silica of cristobalite β
structure which is different from the quartz crystal structure the silica rods are made of indicating
the silica formed condensed onto the silica rod. This deposit will either increase the weight or
make it so brittle that pieces fall off decreasing the weight affecting the results either way.
Sometimes the mass loss and gain may cancel each other out because some deposits stay intact
and some fall off. It was found that if the hole in the lid has a diameter very close to the diameter
of the rod (so as to contain the SiF4 within the crucible) this reaction is reduced leading to more
accurate results.
Errors in this methods are due to:
Variations in experimental measurements; deduced from repetition.
The quartz at the bottom of the rod sometimes chips off because of cracks due to
temperature differences when the silica is removed from the high temperature furnace. A
loss of 0.05 – 0.1g can be expected.
SiF4 reactions can affect the mass measurements especially at low temperatures where
mass loss is low and a difference of 0.1g can have a high percentage change.
Page 56
47
3.2 Density Measurement
The experiment explained in this section was used to measure density of the CR+LiF electrolytes
with varying silica concentrations and temperatures, based on the Archimedean principle.
A graphite crucible was used to contain 300g of CR+LiF melt. The reason for the larger melt size
is that the height of the melt was enough to accommodate an entirely immersed sink without the
walls of the crucible affecting the measured results. Again the fume hood is necessary for this set
up since silica is added to the CR+LiF melt and produces SiF4. The furnace was also purged with
argon between measurements. One whole experiment consisted of a given silica concentration in
the CR+LiF melt while temperature was varied from 920 to 1000°C in 20° intervals and density
measured at each temperature. For each new experiment the concentration of silica (99.5%
purity) was increased in 0.5wt% increments from 0 - 4wt%. The melt was stirred and the
temperature was measured before each density measurement.
A sink, made of nickel ball and held up by a Ni-Cr wire, was used to measure the density of the
molten CR+LiF melt. To avoid errors made from fluctuations in mass due to natural convection
currents the sink was weighed down using a steel rod threaded through the wire. The manner in
which the ball was attached to the wire also ensured that the entire set-up was buoyed as opposed
to only the nickel ball, shown in Figure 22 on page 49.
Before each new experiment the mass of the sink was measured. An estimate was made of how
much of the sink (ball and wire) would be immersed in the CR+LiF melt and this same volume
was immersed in distilled water at room temperature and the apparent mass was measured. The
difference in masses and the known density of distilled water (0.9956 g/cm3 at 24°C) allowed
calculating the density of the sink that would be immersed in the melt according to Archimedes’
principle in Equation (10).
(10)
Density of water was used to calculate density of sink which is assumed to be entirely pure
nickel (even the wires). Volume of the immersed sink at ambient temperature was calculated
Page 57
48
using its measured density and mass. This volume in conjunction with the linear thermal
expansion coefficient of nickel (α = 13.4m.m-1
.K-1
) were used to calculate the volume of the
immersed sink at different temperatures, Equation (11), and hence the densities at each of these
temperatures. This is then used to calculate the density of the CR+LiF melt using Equation (10).
(11)
The sink was suspended from a mass balance which was connected to the computer to record
values every 5 seconds for 30 readings at a given temperature. Figure 23 shows the schematic of
the experimental set up for density measurements. The sink was allowed to come to equilibrium
at the temperature in the melt so as to avoid error due to melt solidification on the sink and
expansion of the sink.
The first density measurement was done at 920°C and the temperature was increased in
increments of 20° from 900 to 1000°C each time a density measurement made. Then the
temperature was decreased in decrements of 20° back to 920°C and again density measurement
was made at each temperature. This procedure was repeated 3 times for each melt composition.
The reason for measuring the density while decreasing the temperature was because of the silica-
cryolite reaction described in Equation (8), which changes the composition of the melt during the
experiment. This means that the density measured while heating the melt might not necessarily
be the same while cooling the melt, since this reaction is both time and temperature dependent.
Furthermore this might lead to variation in the results when the experiment is repeated that may
be mistaken as errors. The results will be further discussed in Chapter 4.
As can be seen in the photograph in Figure 23 the furnace is open and allows for the escape of
SiF4 gas. Using Equation (10) the density of the melt was calculated for varying temperatures
and varying silica concentrations.
Page 58
49
The Ni-Cr wire was twisted around the nickel ball as
to avoid any independent buoyancy of the ball as
opposed to the whole sink; ball, wire and steel rod.
The photograph shows the the sink used for this
method.
Figure 22 Set up of the sink to be immersed in the
CR+LiF melt.
Page 59
50
Errors in this method are due to:
Fluctuations of the mass readings due to natural convection currents in the melt varying
between 0.1-0.4g per 70-80g measured.
Solidification of the melt onto the nickel ball after saturation concentration of silica has
been reached at a particular temperature thereby giving false results (explained in the
results and discussion section).
Slight calculation errors due to considering Ni-Cr wire has the same density and
expansion coefficient as Nickel ball; this error is not as significant as the variations due to
the change in composition and therefore not taken into account. The average mass of the
ball is 21.65g and the average mass of the wire that was wrapped around the ball was
0.75g which is 3% of the whole mass. Calculation error in assuming the density as the
same for both ball and wire will contribute 3% error at most. Whereas the density error
due to compositional change was on average 0.23g/cm3 for 2.25g/cm
3 which is over 9%.
Figure 23 Schematic of the experimental setup for density measurement
Furnace
Mass Balance
Argon Gas Inlet
Electrolyte
Sink
Page 60
51
3.3 Conductivity Measurement
An attempt was made at measuring conductivity of the CR+LiF melt with varying concentration
and temperatures. The four probe electrode had already been used successfully to measure
conductivity for cryolite-silica melts and oxide melts [93, 94]. The set up for this method is
shown in Figure 24.
Figure 24 Schematic of the experimental setup for conductivity measurement
Page 61
52
The resistance of the melt Rmelt is described by equation
Rmelt = (l/A) = (1/σ) (l/A) (12)
Where ρ is resistivity and σ is conductivity. The cell constant is described as l/A and calibration
is needed with a standard solution whose conductivity is already known to determine this
constant. The resistance is measured using electrochemical impedance spectroscopy (EIS) using
a potentiostat and a frequency analyser. An AC voltage is applied at varying frequencies and the
responding current measured. From these values a Nyquist plot is made and the resistance value
Rmeasured is read off the graph where it intersects the real axis (resistance with units of ohm).
Figure 25 shows a typical Nyquist plot when measuring impedance and the inset shows a close
up of where the curve intersects the real axis.
Figure 25 Typical Nyquist plot for impedance spectroscopy
Page 62
53
However R measured = R melt + Rleads + R electrode therefore the Rleads + R electrode should be measured
by short circuiting the electrode with a copper foil and using EIS.
There are drawbacks to this method [95]:
The cell constant of the electrode cell is determined using a solution or melt whose
conductivity is already established in literature. However the cell constant is a function of
temperature and the conductivity of the liquid. To avoid cell constant change due to
expansion of the electrode at higher temperatures, measuring it at working temperatures
using a molten salt is recommended. However to avoid the latter of these errors the
unknown conductivity of the new liquid needs to be similar to that of the liquid used for
calibration.
There are fringe effects when the electrodes are immersed in the melt.
To avoid these errors the Van der Pauw method was attempted [96]. The arrangement of the
electrodes varies as shown in Figure 26. The schematic on the left is the ordinary four probe
method used by Sokhanvaran et al. [93]. The drawing on the right is the Van der Pauw
arrangement.
Figure 26 Two arrangements for four probe electrode
The method uses four electrodes arranged in a square. Again EIS is used and voltage V1 is
applied and current I1 is measured at height H1 in the melt. To make sure any discrepancies are
eliminated due to imperfect square arrangement voltage V2 and I2 are also measured at H1 and
an average resistance RH1 is calculated. This procedure is repeated at another height H2 in the
melt to get RH2; a total of 8 measurements are made for one conductivity value. This is time
Page 63
54
consuming but eliminates fringe effects and the need for cell calibration. The resistivity is
calculated by Equation (11).
(13)
The conductivity, σ, is the inverse of resistivity σ = 1/ρ. However this method also has its
drawbacks in that results are affected by the crucible walls, depth of immersion and the diameter
of the individual probes [96].
This method had been previously used [97] to measure the conductivity of cryolite using
platinum electrodes. Platinum leads to the platinisation effect whereby the values obtained are
inaccurate because of catalytic reactions on the platinum surface leading to solvent
decomposition [98] and increasing apparent conductivity.
The method was modified for the present work for pure cryolite by using molybdenum rods as
electrodes and a voltage of 50mV was applied and the current measured. However an unforeseen
problem was encountered; one of the current electrodes was dissolving up to the length that was
immersed in the melt at such a quick rate that the electrode would lose the immersed length at
the end of one experiment. The only difference between the Van der Pauw method used in this
study and the four probe method used by Sokhanvaran [93] was the arrangement of the
electrodes which leads to conclusion that the electric field created by the electrode arrangement
is the reason for the dissolution of the molybdenum. The following reaction is believed to happen
at the positive counter electrode:
Mo Moz+
+ ze-
The Mo ion goes into the melt. However a deposition of the Mo ion is not seen at any of the
other electrodes. This may be because the electrons at the negative electrode are used to reduce
some other ion in the melt or the deposition of Mo is not noticeable because the electrodes are
slightly oxidising during the experiment and MoO2 created masks any deposition.
This method had to be abandoned because the measurements could not be trusted; the shape and
size of the electrodes were changing during the experiment. The initial results obtained from pure
cryolite are reported below and compared to that obtained by Sokhanvaran et al. in Figure 27.
Page 64
55
There are other results for pure cryolite available but the cryolite used in this study is from the
same batch as used by Sokhanvaran. The measurements stop at 1080°C because the electrode
was completely dissolved and the experiment was not repeated.
Figure 27 Comparing results from different experimental setup
Figure 27 shows that the Van der Pauw results are higher; this could be due to the dimensions of
the electrode changing thus affecting the results.
1.5
2.0
2.5
3.0
3.5
4.0
4.5
1000 1020 1040 1060 1080 1100
Con
du
tivit
y (
S/c
m)
Temperature (°C)
This study - Van der Pauw
Sokhanvaran - four wire [89]
Page 65
56
Chapter 4 Results and Discussions
4.1 Dissolution Behaviour
A typical mass loss versus time graph looks like Figure 28. The graph shows the experimental
values obtained at 1000°C and the theoretical curve using calculated k and CS values. As will be
explained later k and CS values were calculated based on data points up to the 4 hour experiment
time, hence the last data point in Figure 28 does not coincide with the model.
Figure 28 Typical mass loss versus time graph (1000C)
The experimental values were used to fit the data to Equation (7) and minimise sum of errors to
obtain values for mass transfer coefficient k and solubility limit CS. According to Equation (14)
the initial slope of the graph is a function of k, CS and the initial area A of silica rod exposed to
the CR+LiF melt (since CB = 0 initially).
(14)
0
1
2
3
4
5
6
7
8
9
0 1 2 3 4 5 6 7 8
Cu
mu
lati
ve
Mass
Loss
(g)
Time (h)
experimental
model (best fit)
Page 66
57
This is expected since the initial change in mass would increase by increasing the surface area
exposed to the melt. CS, being the “driving force” for the solute to dissolve into the solvent, the
higher the CS value the more mass that can be dissolved. As time tends to infinity the graph tends
to a horizontal asymptote, i.e. dm/dt 0 and Equation (6) would indicate that either CB CS or
A = 0. Since observation indicates that the whole rod is not dissolved then A cannot equal 0
which means that CB approaches CS. Figure 29 shows the results calculated for mass transfer
coefficient k for experiments done at various temperatures and it increases with increasing
temperature.
Figure 29 k versus temperature
The values of k were fitted to an exponential curve because k is a function of D (diffusivity):
and D is an exponential function of temperature
. n can vary between 0.5
and 1 depending on the chemicals. n = 0.5 when the diffusion coefficient or time of exposure of
solute in solvent is small; fluid comes in contact with the solid, mixes and then moves away. n =
1 when diffusion coefficient or time of exposure is large and transfer of solid is through a
diffusion layer. In realistic experiments n is a value between 0.5 and 1 and for this study it is
more likely that n is closer to 0.5 because of the slow dissolution rate (hence k value) and small
time of exposure due to high convection currents at high temperatures.
0
1
2
3
4
5
6
7
8
9
10
900 920 940 960 980 1000 1020
Mass
Tra
nsf
er C
oef
fici
ent
(cm
/h)
Temperature (°C)
Page 67
58
As is expected from the relation between k and D, the mass transfer coefficient increases with
increasing temperature. Diffusivity is difficult to measure directly through the standard
electrochemical process because silicon is speculated to form a large complex molecule in the
melt and the mobility is very small affecting the measurements. Attempts at measuring
diffusivity were abandoned to study dissolution behaviour. However it is necessary to point out
that viscosity affects dissolution rate and therefore the apparent mass transfer coefficient. Since
viscosity changes with temperature, usually decreasing as temperature increases, and
composition as the silica dissolves into the melt, viscosity should be measured to determine the
exact relation with mass transfer coefficient values.
Figure 30 shows the results obtained for solubility limit CS (also calculated with data points up to
4 hour result) which increases with temperature and has a linear relationship between 920 and
1000°C. The line was constructed using least squares regression. The use of the solubility limit is
twofold; 1) to avoid the anode effect from lack of silica or too many inclusions in the deposit
from too much silica, 2) to construct part of a ternary phase diagram; the liquidus line.
Figure 30 CS versus temperature
1.0
1.5
2.0
2.5
3.0
3.5
4.0
900 920 940 960 980 1000 1020
Solu
bil
ity L
imit
(w
t%)
Temperature (C)
Page 68
59
The dissolution behaviour experiment was conducted for pure cryolite to validate the
experimental procedure. The solubility limit in pure cryolite was calculated, using data points up
to the 8 hour result, and found to be varying between 9-15wt% which is significantly higher than
the value deduced by Grjotheim et al. [38], 5wt% at 1020°C. As explained in Chapter 3 the
EPMA analysis indicated a formation of an oxide phase containing silicon, aluminium and
sodium. After the silica dissolves in the cryolite, once it has reached saturation it most likely
continues to react with the cryolite to form the high silica concentration phases. This may
explain why, when the dissolution behaviour experiment is conducted, the solubility values are
high; the melt becomes saturated with silicon and then compounds of albite, jadeite and
nepheline (based on the duration of the reaction) are formed and precipitate out.
To determine how the time affected the calculated k and CS values, both sets of data were
calculated based on all 6 data points (up to 8 hours) and only 5 data points (up to 4 hours) as
shown in Table 8.
Table 8 Effect of time on the calculated k and Cs
With 8 hour point With 4 hour point
Temperature
(°C)
k
(cm/h)
CS
(wt%)
k
(cm/h)
CS
(wt%)
CR+
13.4wt%LiF
920 2.7 2.2 2.7 2.2
940 4.0 2.2 4.3 2.0
960 4.8 2.8 5.0 2.7
980 5.3 3.2 6.5 2.8
1000 7.7 3.8 9.1 3.5
Pure cryolite 1025 2.3 11.1 4.5 7.5
Generally the percentage change of k and CS values are not significant for the melts containing
lithium fluoride whereas for the pure cryolite melt the percentage change is significant. The
solubility limit has decreased from 11.1 to 7.5wt% which closer reflects the value found by
Grjotheim et al. This seems to indicate that after a while the silicon reacts with the cryolite as
opposed to dissolve in it and the lithium fluoride acts as an inhibitor to this reaction because it
decreases the overall solubility of silica in cryolite. This is a reasonable assumption as lithium
Page 69
60
fluoride makes up 56 mol% of the melt and might be attributed to silicon complex formation
affected by LiF addition. This was also the reason why when calculating k and CS values only
data points up to the 4 hour result was used; the longer the experiment the more likely the silica
is reacting with the cryolite as opposed to dissolving in it.
It is reported that the addition of LiF decreases the solubility limit of alumina in cryolite and
reduces the dissolution rate [28]. A lower solubility limit might explain why the dissolution
decreases; there is a lower driving force for the alumina to dissolve in the cryolite. Section 2.4 in
Chapter 2 tried to explain why the solubility of alumina would decrease with addition of LiF;
increasing basicity of the melt decreases solubility limit of alumina in cryolite. In this study
addition of LiF decreases the solubility limit of silica in cryolite but increases the dissolution
rate. It is likely that the same mechanism for alumina would explain why the solubility of silica
is reduced with addition of LiF, however the increased dissolution rate might be explained by
examining the mass transfer coefficients.
When comparing mass transfer coefficients the result for pure cryolite was 4.5cm/h at 1025°C
and for CR+LiF melt it was 9.1cm/h at 1000°C. Either this increase in mass transfer is due to the
increased superheat for the CR+LiF melt (1000°C compared to the 15°C for pure cryolite) or it is
because the lithium fluoride increases dissolution rate of silica into the melt. Although at 920°C
the dissolution rate is only 2.7 for the CR+LiF melt this could be due to the lower working
temperature range thus affecting the dissolution kinetics of silica. LiF is reported in literature as
decreasing viscosity of melts; for pure cryolite at 1025°C the viscosity is reported as 2.16mPa.s
whereas at 15wt%LiF in cryolite at 1000°C the viscosity is about 1.7mPa.s, [33]. This decrease
in viscosity will might allow more convection flow in the melt and hence more dissolution. The
effect of viscosity and temperature go together and cannot be distinguished without further
experiments. It is also likely that the Si-O chains are terminated with Li+ ions with more
thermodynamic favourability and hence encourage the dissolution of SiO2 into the melt.
Furthermore if a reaction happens between the solute and solvent, as is applicable to this case,
then the apparent mass transfer coefficient is increased. For now it can be concluded that lithium
fluoride increases the apparent mass transfer coefficient.
Page 70
61
It should be noted that the experiments with the largest variation when looking at the mass loss
versus time curves were those done at lower temperatures as shown in Figure 31.
Figure 31 Comparison of variation of mass loss for different temperatures
At lower temperatures the natural convection flows are slower and might not average out over
time as it would for higher temperatures hence the dissolution of the silica can be uneven leading
to an uneven dissolution curve. Also as the silica reacts with the melt changing the composition
and hence the melting temperature of the now changed melt, this effect may affect the results
more at lower temperatures. From this it can be concluded that this method is more accurate the
higher the superheat of the melt.
0
1
2
3
4
5
6
7
8
9
0 2 4 6 8
Cu
mu
lati
ve
Ma
ss L
oss
(g
)
Time (h)
1000 - exp 1
1000 - exp 2
1000 - exp 3
920 - exp 1
920 - exp 2
920 - exp 3
Page 71
62
4.2 Density
The results for density measurements for each temperature with varying silica concentrations are
presented in Figure 32. The data showed a considerable scatter particularly with respect to
temperature. However, for each temperature, with addition of silica the density first increased,
and then when the silica content reached ≈ 3wt%, began to decline.
The density increase in the first part of each curve may be related to the increased silica
concentration because silica has a higher density than the silica-free electrolyte. The decline in
the second part is most likely because the electrolyte has reached saturation and when the nickel
ball is immersed there is precipitation of solid phases on the ball which gives false results. The
actual volume of the sink will be larger than the assumed “clean” sink, resulting in a smaller
calculated density. A similar trend was seen in a parallel work on pure cryolite [54]. The value at
which this phenomenon is observed corresponds to the solubility limits obtained in Section 4.1
for the dissolution experiments.
LiF addition results in a decrease in silica solubility and generally increases density of the melt
with silica addition compared to pure cryolite as can be seen in Figure 33. This graph shows one
set of data conducted at 1010C for pure cryolite, study done by Sokhanvaran [54], and 1000C
for CR+LiF used in this study. This is the opposite behaviour to alumina addition as shown in
Table 6; LiF addition to cryolite – alumina melt decreases density. A comparison of varying
alumina concentration in a fixed ratio of 13.8wt%LiF-Cryolite [28] is made with the melt
composition in this study in Figure 34. Figure 35 (page 65) shows a compilation of all the data of
pure cryolite with oxide and CR+LiF with oxide. With increasing alumina content the density
decreases but with increasing silica content the density increases. Addition of 14wt% LiF
decreases the overall density of cryolite-alumina melt whereas it increases the overall density of
cryolite-silica melt. This indicates one cannot make assumptions for silica using alumina data
and hence further studies on other properties are necessary.
Page 72
63
Figure 32 Density versus silica concentration for five temperatures
1.5
1.7
1.9
2.1
2.3
2.5
2.7
0 1 2 3 4
Den
sity
(g
/cm
3)
Silica Concentration (wt%)
920°C
940°C
960°C
980°C
1000°C
1.5
1.7
1.9
2.1
2.3
2.5
2.7
0 1 2 3 4
Den
sity
(g/c
m3)
Silica Concentration (wt%)
920°C
1.5
1.7
1.9
2.1
2.3
2.5
2.7
0 1 2 3 4
Den
sity
(g/c
m3)
Silica Concentration (wt%)
940°C
1.5
1.7
1.9
2.1
2.3
2.5
2.7
0 1 2 3 4
Den
sity
(g
/cm
3)
Silica Concentration (wt%)
960°C
1.5
1.7
1.9
2.1
2.3
2.5
2.7
0 1 2 3 4
Den
sity
(g
/cm
3)
Silica Concentration (wt%)
980°C
1.5
1.7
1.9
2.1
2.3
2.5
2.7
0 1 2 3 4
Den
sity
(g/c
m3)
Silica Concentration (wt%)
1000°C
Page 73
64
Figure 33 Density comparison with and without LiF at ~1000°C
Figure 34 Density comparison of silica versus alumina addition at 1000°C
Molar volume was calculated from the density of the various constituents of the melt (as shown
in Table 9). The molar volume of LiF is lower than that of cryolite therefore if it made up 13wt%
of the melt then the density should increase when compared to pure cryolite. The difference in
the molar volume of alumina and silica is very small so the reason why when LiF is added to the
melt it would have opposite effects on the densities would be down to the complexes formed in
the melt.
1.9
2
2.1
2.2
2.3
2.4
2.5
0 1 2 3 4 5 6 7
Den
sity
(g/c
m3)
Silica Concentration (wt%)
Sokhanvaran (1010°C)
without LiF [54]
This study (1000°C)
with LiF
1.9
2.0
2.1
2.2
2.3
2.4
2.5
0 1 2 3 4 5 6 7 8 9
Den
sity
(g/c
m3)
Oxide Concentration (wt%)
Matiasovsky (1000°C)
with Alumina [28]
This study (1000°C)
with Silica
Page 74
65
Table 9 Molar volumes of the various constituents of the melt
Molecular Mass
(g/moles)
Density
(g/cm3)
Molar volume
(cm3/mole)
Na3AlF6 209.94 2.04 [54] 102.91
LiF 25.94 1.74 [99] 14.95
SiO2 60.08 2.20* 27.31
Al2O3 101.96 3.69** 27.63
Aluminium is reported to form Al2OFx y-
complex ions in the melt [28, 33]. Addition of
aluminium oxide would provide the melt with a source of aluminium to form these large
complexes. It was also concluded from previous studies that increasing the cryolite ratio would
increase the coordination number of the complexes [33]. Addition of LiF would effectively
increase the mole ratio (from 3 to 8 with 13.8wt%LiF addition) as it could be redefined as
(NaF+LiF)/AlF3. Both these factors might explain why in the cryolite-alumina melt addition of
lithium fluoride would decrease the density. And an increase in density of the cryolite-silica melt
would suggest the opposite behaviour hence a detailed study of silica complex formation in
cryolite and CR+LiF melts need to be conducted.
Figure 35 Comparison of effect of LiF addition with different oxides
1.90
2.00
2.10
2.20
2.30
2.40
2.50
0 2 4 6 8 10 12
Den
sity
(g/c
m3)
Oxide Concentration (wt%)
Grjotheim (1000°C) without LiF with Alumina [45]
Sokhanvaran (1010°C) without LiF with Silica [54]
Matiasovsky (1000°C) with LiF with Alumina [28]
This study (1000°C) with LiF with Silica
* http://www.alfa.com/en/gp100w.pgm?dsstk=013024&rnd=394767240 ** http://accuratus.com/alumox.html
Page 75
66
When looking at the density behaviour with temperature as shown in Figure 36, the trend is not
as expected i.e. a decrease in density with an increase in temperature. In fact, while increasing
temperature from 920 to 1000°C the density increases and then decreases. Conversely when
temperature is reduced from 1000 to 920°C the density increases with decreasing temperature
which is expected.
Figure 36 Density versus temperature for 1wt% SiO2 electrolyte
This behaviour is due to the reaction as described in Equation (8) (stated earlier):
SiO2 + Na3AlF6 = xNa2O.yAl2O3.zSiO2 + NaF + SiF4↑
implying that the melt composition is changing during the experiment. The composition is a
function of time and temperature. It is also an interesting fact that according to Snow et al. [91]
even the x, y, and z values of the compound is dependent on time, forming at first jadeite, then
disproportioning into albite and nepheline. While heating the melt the reaction is continually
happening, the rate of reaction depends on the temperature and the extent to which the reaction
has happened depends on the time. Hence when repeating the experiment, reproducible results
cannot be achieved. While cooling the melt the reaction is in equilibrium and the expected
behaviour of density increase with decreasing temperature is observed. What these results do
2.0
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
900 920 940 960 980 1000
Den
isty
(g/c
m3)
Temperature (°C)
1% silica - exp 1 - heating
1% silica - exp 2 - heating
1% silica - exp 3 - heating
1% silica - exp 4 - cooling
Page 76
67
indicate is the maximum density that can be expected at a given temperature and silica
concentration. The electrolysis process can then be designed to accommodate for this. An
additional implication of the result is that the molten (Na, Si, Al)xOy compound is denser than
CR+LiF and further addition of silica will change the composition of the melt to such an extent
that cryolite may be entirely tied up in forming (Na, Si, Al)xOy compounds.
Figure 37 Variations of density with silica concentration versus temperature
2.0
2.1
2.2
2.3
2.4
2.5
900 920 940 960 980 1000 1020
Den
sity
(g/c
m3)
Temperature (°C)
0%
0.50%
1%
1.50%
2%
2.50%
3%
Page 77
68
Due to a combination of the silica-cryolite reaction and varying silica concentration, the effect of
temperature is not observed in Figure 37 and average density is established for varying silica
concentration. If aluminium was used as a liquid cathode then the cathode would start with about
2.3g/cm3. As shown in Figure 38, the solid line indicates the density of liquid aluminium. The
dotted line indicates the maximum concentration of silica allowed, to make sure the melt will
float on top of the aluminium cathode. As the silicon dissolves into the cathode the density will
increase [100], it is possible to use up to 2.5wt% silica in the melt and make sure the melt will
float on top of the aluminium cathode. Other molten cathode materials like copper and tin have
high enough densities to stay below the molten CR+LiF melt.
Figure 38 Density compared to liquid aluminium
1.50
1.70
1.90
2.10
2.30
2.50
2.70
0 1 2 3 4
Den
sity
(g/c
m3)
Silica Concentration (wt%)
920°C
940°C
960°C
980°C
1000°C
Page 78
69
Chapter 5 Conclusions and Future Work
5.1 Summary and Conclusions
Three properties were measured for a Cryolite-13.8wt%LiF melt with varying concentrations of
silica at varying temperatures; solubility limit, dissolution rate and density. Two experiments
were conducted a) dissolution behaviour which gave solubility limit and dissolution rate and b)
density measurement using Archimedes’ principle. Silica concentrations were varied at 0.5%
intervals between 0 and 4% and the temperature at 20°C intervals between 900 and 1000°C.
The dissolution behaviour experiment results indicate that solubility is in the range of 2-4wt%
and increases with a linear trend with increasing temperature in the range of 900-1000°C. The
mass transfer coefficient increases exponentially with temperature, ranging from 3-9cm/h. The
solubility values are used to a) prevent depletion of silicon ions at the cathode and creating the
anode effect where a film of gas forms around the anode preventing reduction of silicon ions and
b) avoid adding too much silica which will settle to the bottom be inclusions in the liquid
cathode. The solubility limits can also be used to construct a phase diagram with fixed cryolite to
lithium fluoride ratio. This study concluded that lithium fluoride increases the dissolution rate
but decreases the solubility limit of silica into cryolite. Also once the silica saturates the melt it
continues to react with the melt and lithium fluoride hinders this reaction.
The Archimedes’ experiments illustrated that the density of the melt does not show the common
trend of decrease in density with increase in temperature. When heating from 920 to 1000°C the
density increases and then decreases, a maximum appearing around 960°C. This indicates a
reaction is taking place which changes the composition of the melt. There is a release of SiF4
when the silica reacts with the cryolite-lithium fluoride melt. However after the reaction has
reached equilibrium the density increases with decreasing temperature (cooling from 1000 to
920°C). The reaction and silica concentration counters each other such that the effect of
temperature is not observed.
Page 79
70
The density values indicate if the liquid alloy of cathode and silicon will float on top or sink
below the electrolyte melt. Up to 2.5wt% of silica dissolved in the electrolyte melt can ensure
that an aluminium-silicon alloy will sink below the electrolyte. Addition of lithium fluoride to
the cryolite-silica melt increases the density; the opposite behaviour to cryolite-alumina melt.
Study should be conducted on the complex formation of silica in cryolite and lithium fluoride
melts to gain a deeper and more fundamental understanding of the melt structure. The decrease
in density for the Al2O3-Cryolite melt is reasoned with alumina forming complexes with oxygen
and fluoride ions in the cryolite melt [33]. Sokhanvaran [54] has found the cryolite-silica melt
increases with silica content; addition of LiF increases the density further which can be attributed
to the lower molar volume. Certainly this study has shown that information gleaned from
alumina-cryolite-lithium fluoride melts (summarised in [33]) cannot always be extrapolated to
make assumptions about silica-cryolite-lithium fluoride melts. The table below summarises how
the oxides affect the CR+LiF melt differently (or to put it another way; how a fixed wt% addition
of LiF affects the CR+oxide melt).
CR+LiF Melt Alumina (Al2O3) Silica (SiO2)
Dissolution Rate
Addition of LiF to cryolite
decreases the dissolution of
alumina into the melt.
Addition of LiF to cryolite
increases the dissolution of silica
into the melt.
Viscosity
Although addition of LiF
decreases viscosity of cryolite, it
increases with increasing
alumina concentration.
Unknown
Solubility Limit
Addition of LiF to cryolite
decreases the solubility of
alumina in the melt.
Addition of LiF to cryolite
decreases the solubility of silica
in the melt.
Density
Increasing alumina concentration
decreases the density. Addition
of LiF decreases the density.
Increasing silica concentration
increases the density. Addition of
LiF increases the density.
Page 80
71
5.2 Future Works
There are several studies that can be done to continue the investigation on the cryolite-lithium
fluoride-silica melts.
Investigating further the dissolution behaviour; measure mass loss at shorter time intervals
between the 2 and 8 hour points, i.e. measure every hour or even half hour to see if a cross
over between dissolution of silica and reaction of silica with cryolite can be identified.
Improve upon density measurements by developing a method to continuously measure
density to examine how it varies with time at a specific temperature and repeat this
procedure for different temperature to get a comprehensive set of results.
Use the four electrode method to measure conductivity; this is a procedure that has already
been used to measure conductivity of cryolite – silica melts [93].
Measure other properties such as viscosity and surface tension; several methods are
available and further studies will need to be done to determine which method will be most
suitable. The final method will depend on the materials the equipment will have to be made
of to withstand working temperatures and the corrosive cryolite.
Raman spectroscopy to study complex formation for deeper understanding of melt structure.
Attempt actual electrolysis to obtain silicon alloy.
Page 81
72
References
[1] H. J. Moller, Semiconductors for Solar Cells Applications: Pergamon Press, 1991.
[2] M. G. Mauk, "Silicon Solar Cells: Physical Metallurgy Principles," Journal of Materials,
vol. 55, pp. 38-42, 2003.
[3] GreenZu, "Breakdown of Solar Panels Cost." vol. 2012, 2012.
[4] D. Elwell and G. M. Rao, "Electrolytic Production of Silicon," Journal of Applied
Electrochemistry, vol. 18, pp. 15-22, 1988.
[5] D. Elwell and R. S. Feigelson, "Electrodeposition of Solar Silicon," Solar Energy
Materials, vol. 6, pp. 123-145, 1981.
[6] R. Monnier, "L’obtention et le Raffinage du Silicium par Voie Electrochimique," Chimia,
vol. 37, pp. 109-124, 1983.
[7] U. S. EIA, "Energy Explained, Your Guide to Understanding Energy," 2012.
[8] C. Flavin, "Low-Carbon Energy: A Roadmap," Worldwatch Institute, Washington,
DC2008.
[9] B. R. Bathey and M. C. Cretella, "Review Solar-grade Silicon," Journal of Materials
Science, vol. 17, pp. 3077-3096, 1982.
[10] G. Flamant, V. Kurtcuoglu, J. Murray, and A. Sieinfeld, "Purification of Metallurgical
Grade Silicon by a Solar Process," Solar Energy Materials and Solar Cells, vol. 90, pp.
2099-2106, 2006.
[11] K. Morita and T. Miki, "Thermodynamics of Solar-grade Silicon Refining,"
Intermetallics, vol. 11, pp. 1111–1117, 2003.
[12] J. L. Gumaste, B. C. Mohanty, R. K. Galgali, U. Syamaprasad, B. B. Nayak, S. K. Singh,
and P. K. Jena, "Solvent Refining of Metallurgical Grade Silicon," Solar Energy
Materials, vol. 16, pp. 289-296, 1987.
[13] S. Esfahani, "Solvent Refining of Metallurgical Grade Silicon Using Iron," in Materials
Science and Engineering. vol. MASc: University of Toronto, 2010.
[14] T. Y. K. Morita, "Refining of Silicon during its Solidification from a Si–Al Melt,"
Journal of Crystal Growth, vol. 311, pp. 776-779, 2009.
[15] A. E. Austin, "Silicon Electrodeposition." vol. US Patent 3990953 USA: Battelle
Development Corporation, 1976.
[16] B. G. Gribov and K. V. Zinov'ev, "Preparation of High-Purity Silicon for Solar Cells,"
Inorganic Materials, vol. 39, pp. 653-662, 2003.
[17] Q. Ma, W. Liu, B. Wang, and Q. Meng, "Electrodeposition of Silicon in Organic Solvent
containing Silicon Chloride," Advanced Materials Research, vol. 79-82, pp. 1635-8,
2009.
[18] R. Monnier and D. Barakat, "Dual Cell Refining of Silicon and Germanium." vol. US
Patent 3219561 USA: Gen Trustee Company Inc 1962.
[19] G. M. Rao, D. Elwell, and R. S. Feigelson, "Electrowinning of Silicon from K2SiF6-
Molten Fluoride Systems," Journal of Electrochemical Society: Electrochemical Science
and Technology, vol. 127, pp. 1940-1944, 1980.
[20] D. Elwell, "Electrowinning of Silicon from Solutions of Silica in Alkali Metal
Fluoride/Alkaline Earth Fluoride Eutectics," Solar Energy Materials, vol. 5, pp. 205-210,
1981.
Page 82
73
[21] G. M. Rao, D. Elwell, R. S. Feigelson, "Electrodeposition of Silicon onto Graphite,"
Journal of Electrochemical Society: Electrochemical Science and Technology, vol. 128,
pp. 1708-1711, 1981.
[22] R. C. De Mattei, D. Elwell, and R. S. Feigelson, "Electrodeposition of Silicon at
Temperatures above its Melting Point," Journal of Electrochemical Society:
Electrochemical Science and Technology, vol. 128, pp. 1712-1714, 1981.
[23] R. Monnier, P. Grandjean, J. Zahler, J. L. Blanc, and T. Tola, "Recherches sur
l’obtention, la séparation et la purification du tantale et du niobium, notamment par voie
électrolytique III. Obtention du tantale, du niobium et du vanadium par électrolyse des
solutions cryolithiques de leurs pentoxydes," Helvetica Chimica Acta, vol. 44, pp. 1862-
1867, 1961.
[24] R. Monnier, P. Grandjean, and J. Zahler, "Recherches sur l’obtention, la séparation et la
purification du tantale et du niobium, notamment par voie électrolytique IV. Essais de
séparation électrolytique du niobium et du tantale," Helvetica Chimica Acta, vol. 46, pp.
2966-2970, 1963.
[25] T. Oishi, M. Watanabe, K. Koyama, M. Tanaka, and K. Saegusa, "Process for Solar
Grade Silicon Production by Molten Salt Electrolysis Using Aluminum-Silicon Liquid
Alloy," Journal of The Electrochemical Society, vol. 158, pp. E93-E99, 2011.
[26] R. H. Hopkins, J. R. Davis, A. Rohatgi, R. B. Campbell, P. D. Blais, P. Rai-Choudhury,
R. E. Stapleton, H. C. Mollenkopf, and J. R. McCormick, "Effect of Impurities and
Processing on Silicon Solar Cells " Westinghouse Research & Development Center and
Dow Coming Corporation1980.
[27] P. D. Desai, "Thermodynamic Properties of Iron and Silicon " Journal of Physical and
Chemical Reference Data, vol. 15, pp. 967-983, 1986.
[28] K. Grjotheim, C. Krohn, M. Malinovsky, K. Matiašovský, and J. Thonstad, Aluminium
Electrolysis Fundimentals of Hall-Heroult Process, 2nd ed. Dusseldorf: Aluminium
Verlag Dusseldorf, 1982.
[29] M. J. Assael, K. Kakosimos, R. M. Banish, J. Brillo, I. Egry, R. Brooks, P. N. Quested,
K. C. Mills, A. Nagashima, Y. Sato, and W. A. Wakeham, "Reference Data for the
Density and Viscosity of Liquid Aluminum and Liquid Iron," Journal of Physical and
Chemical Reference Data, vol. 35, pp. 285-300, 2006.
[30] P. A. Foster, "Melting Point of Cryolite," Journal of the American Ceramic Society, vol.
51, pp. 107-109, 1968.
[31] J. Brynestad, K. Grjotheim, F. Gronvold, J. L. Holm, and S. Urnes, "Cryolite + Alumina
Phase Diagram and the Constitution of Melt in this System," Discussions of the Faraday
Society, vol. 32, pp. 90-96, 1961.
[32] J. L. Holm and B. J. Holm, "Phase Relations and Thermodynamic Properties in the
Ternary Reciprocal System LiF-NaF-Na3AlF6-Li3AlF6," Thermochimica Acta, vol. 6, pp.
375-398, 1973.
[33] G. Mamantov and R. Marassi, Molten Salt Chemistry: An Introduction and Selected
Applications: D. Reidel, 1987.
[34] Y. Zhang and R. A. Rapp, "Modeling the Dependence of Alumina Solubility on
Temperature and Melt Composition in Cryolite- Based Melts," Metallurgical and
Materials Transactions B, vol. 35, pp. 509-515, 2004.
[35] K. Hellberg and J. Massonne, "Process for making Cryolite ". vol. US Patent 3875291
USA: Kali-Chemie Aktiengesellschaft (Hannover, DT) 1975.
Page 83
74
[36] J. Aigueperse and e. al., "Fluorine Compounds, Inorganic," in Ullman's Encyclopedia of
Industrial Compounds. vol. 15, W. Gerhartz and e. al, Eds.: VCH Publishers, 1985.
[37] K. Grjotheim, K. Matiašovský, and P. Fellner, "The Electrodeposition of Silicon from
Cryolite Melts," in Light Metals 1982 Dallas, Texas, 1982.
[38] R. C. DeMattei and R. S. Feigelson, "Electrochemical Deposition of Semiconductors," in
Electrochemistry of Semiconductors and Electronics, J. M. F. Ludwig, Ed.: William
Andrew Publishing/Noyes, 1992.
[39] J. Ming, L. Yan-Qing, T. Zhong-Liang, L. Fang-Yang, L. Jie, X. Peng-Fei, and L. Ye-
Xiang, "Electrodeposition Behavior of Silicon from Na3AlF6-LiF Melts," Acta Physico-
Chimica Sinica, vol. 27, pp. 1108-1115, 2011.
[40] D. F. Weill and W. J. Fyfe, "The 1010° and 800°C Isothermal Sections in the System
Na3AIF6-AI2O3-SiO2," Journal of Electrochemical Society, vol. 111, pp. 582-585, 1964.
[41] R. G. Haverkamp and B. J. Welch, "Modelling the dissolution of alumina powder in
cryolite," Chemical Engineering and Processing, vol. 37, pp. 177–187, 1998.
[42] N. V. Vasyunina, I. P. Vasyunina, Y. G. Mikhalev, and A. M. Vinogradov, "The
Solubility and Dissolution Rate of Alumina in Acidic Cryolite Aluminous Melts,"
Russian Journal of Non-Ferrous Metals, vol. 50, pp. 338–342, 2009.
[43] D. R. Poirier and G. H. Geiger, Transport Phenomena in Materials Processing: Wiley,
1998.
[44] C. Nicollin, J. Lenoir, and C. Eyraud, Bulletin de la Société Chimique de France, vol. 2,
1967.
[45] K. Grjotheim, K. Matiašovský, P. Fellner, and A. Silný, "Electrolytic Deposition of
Silicon and of Silicon alloys Part I: Physiochemical Properties of the Na3AlF6-Al2O3-
SiO2 Mixtures," Canadian Metallurgical Quarterly, vol. 10, pp. 79-82, 1971.
[46] Y. Zhang, X. Wu, and R. A. Rapp, "Solubility of Alumina in Cryolite Melts:
Measurements and Modeling at 1300 K," Metallurgical and Materials Transactions B,
vol. 34B, pp. 235-242, 2003.
[47] F. J. Frere, "Equilibrium in Fluoride Systems. I. Solubility of Cryolite in Aqueous
Solutions of Iron and Aluminum Salts at 25°C," Journal of the American Chemical
Society, vol. 58, pp. 1695–1697, 1936.
[48] T. E. Jentoftsen, O. A. Lorensten, E. W. Dewing, G. M. Haarberg, and J. Thonstad,
"Solubility of Some Transition Metal Oxides in Cryolite-Alumina Melts: Part I.
Solubility of FeO, FeAl2O4, NiO, and NiAl2O4," Metallurgical and Materials
Transactions B, vol. 33B, pp. 901-908, 2002.
[49] T. E. Jentoftsen, O. A. Lorensten, E. W. Dewing, G. M. Haarberg, and J. Thonstad,
"Solubility of Some Transition Metal Oxides in Cryolite-Alumina Melts: Part II.
Solubility of TiO2," Metallurgical and Materials Transactions B, vol. 33B, pp. 909-913,
2002.
[50] T. E. Jentoftsen, O. A. Lorensten, E. W. Dewing, G. M. Haarberg, and J. Thonstad, "The
Solubility of Some Transition Metal Oxides in Cryolite-Alumina Melts: Part III.
Solubility of CuO and Cu2O," Metallurgical and Materials Transactions B, vol. 38B, pp.
833-839, 2007.
[51] P. Kubíček and T. Pepřica, "Diffusion in Molten Metals and Melts: Application to
Diffusion in Molten Iron," International Metals Reviews, vol. 28, pp. 131-157, 1983.
[52] K. Grjotheim, T. Naterstad, K. E. Torklep, and H. A. Øye, "Tracer Diffusion Coefficients
Molten Salts I. Methods," Electrochimica Acta, vol. 23, 1978.
Page 84
75
[53] R. A. Khairulin, S. V. Stankus, R. N. Abdullaev, Y. A. Plevachuk, and K. Y. Shunyaev,
"The Density and the Binary Diffusion Coefficients of Silver-Tin Melts," Thermophysics
and Aeromechanics, vol. 17, pp. 391-396, 2010.
[54] S. Sokhanvaran, M. Barati, S. Thomas, "Physiochemical Properties of Cryolite-Silica
Melts," in COM 2011 Montreal, QC, MetSoc, 2011, pp. 25-34.
[55] E. N. Lyakhovich, E. D. Shcherbakov, L. G. Voroshrim, and G. G. Panich, in Vestsi
Akademii Navuk, Belarus S. S. R., 1971.
[56] S. Deville, Comptes rendus des séances de l'Académie des sciences, vol. 39, p. 323, 1854.
[57] F. Ullik, Sitzungsberichte der Kaiserlichen Akademie der Wissenschaften in Wien, vol.
52, p. 115, 1865.
[58] G. Gore, "On the Electrodeposition of Carbon," Chemical News and Journal of Industrial
Science, vol. 50, pp. 113-115, 1884.
[59] W. Hampe, Chemiker Zeitung, vol. 12, p. 115, 1888.
[60] M. Dodero, "Electrolytic Preparation of Calcium Silicide," Comptes rendus des séances
de l'Académie des sciences, vol. 109, p. 566, 1934.
[61] M. Dodero, "Electrolysis of Molten Alkali and Alkaline Earth Silicates," Bulletin de la
Société Chimique de France, vol. 6, p. 209, 1939.
[62] H. V. Wartenberg, "Über Silicium," Zeitschrift für anorganische und allgemeine Chemie,
vol. 265, pp. 186-200, 1951.
[63] D. R. Stern and Q. H. McKenna, "Production of Pure Elemental Silicon." vol. US Patent
2892763 USA: American Potash & Chem Corp, 1959.
[64] F. Olstowski, "Preparation of Silicon." vol. US Patent 3022233 USA: DOW Chemical
Co. , 1962.
[65] P. J. Jorgensen, "Electrolysis of SiO2 on Silicon," The Journal of Chemical Physics, vol.
49, pp. 1594-1598, 1968.
[66] U. Cohen, "Epitaxial Growth of Silicon or Germanium by Electrodeposition from Molten
Salts ". vol. US 3983012 USA: The Board of Trustees of Leland Stanford Junior
University, 1975.
[67] U. Cohen and R. A. Huggins, "Silicon Epitaxial Growth by Electrodeposition from
Molten Fluorides," Journal of Electrochemical Society: Electrochemical Science and
Technology, vol. 123, 1976.
[68] U. Cohen, "Some Prospective Applications of Silicon Electrodeposition from Molten
Fluorides to Solar Cell Fabrication," Journal of Electronic Materials, vol. 6, pp. 607-643,
1977.
[69] R. D. Rauh, T. L. Rose, S. H. White, R. A. Boudreau, T. O. Hoover, D. L. Natwig, "Final
Report," EIC Corporation, USA1981.
[70] J. M. Olson and K. L. Carleton, "A Semipermeable Anode for Silicon Electrorefining,"
Journal of The Electrochemical Society, vol. 128, pp. 2698-2699, 1981.
[71] K. L. Carleton, J. M. Olson, and A. Kibbler, "Electrochemical Nucleation and Growth of
Silicon in Molten Fluorides," Journal of Electrochemical Society: Electrochemical
Science and Technology, vol. 130, pp. 782-786, 1983.
[72] J. M. Olson and A. Kibbler, "Electrowinning of Silicon Using a Molten Tin Cathode," in
Electrochemical Society, 160th Fall Meeting, Extended Abstracts, USA, 1981, p. 1130.
[73] J. R. Stubergh and Z. Liu, "Preparation of Pure Silicon by Electrowinning in a Bytownite-
Cryolite Melt," Metallurgical and Materials Transactions, vol. 27B, pp. 895-900, 1996.
Page 85
76
[74] K. Yasuda, T. Nohira, Y. H. Ogata, and Y. Ito, "Direct Electrolytic Reduction of Solid
Silicon Dioxide in Molten LiCl–KCl–CaCl2 at 773 K," Journal of The Electrochemical
Society, vol. 152, pp. D208-D212, 2005.
[75] A. Minet, "Électrolyse par fusion ignée des sels de bore et du silicium," Comptes rendus
des séances de l'Académie des sciences, vol. 112, p. 1215, 1891.
[76] H. N. Warren, Chemical News, vol. 67, p. 303, 1893.
[77] K. P. Batashev and A. Y. Zhurdin, Metallurg, vol. 2, pp. 66-77, 1933.
[78] R. Monnier and D. Barakat, "Contribution à l’étude du Comportement de la Silice dans
les Bains de Cryolithe Fondue," Helvetica Chimica Acta, vol. 40, pp. 2041-2045, 1957.
[79] R. Monnier and P. Grandjean, "Recherches sur l’obtention, la séparation et la purification
du tantale et du niobium, notamment par voie électrolytique II. Etude d’une méthode de
détermination des tensions effectives d’éctrolyse et des tensions pratiques de
décomposition; application de cette méthode à des solutions cryolithiques d’oxydes,"
Helvetica Chimica Acta, vol. 43, pp. 2163-2172, 1960.
[80] R. Monnier, D. Barakat, and J. C. Giacometti, "Refining of Silicon and Germanium ".
vol. US Patent 3254010: Gen Trustee Company Inc., 1962.
[81] R. Monnier and J. C. Giacometti, "Recherches sur le raffinage éctrolytique du silicium,"
Helvetica Chimica Acta, vol. 47, pp. 345-353, 1964.
[82] K. Grjotheim and K. Matiašovský, "Preparation of Si-Al Alloys by Electrolysis of Molten
Cryolite," Hutnické Listy, vol. 25, pp. 876-882, 1970.
[83] K. Grjotheim, K. Matiašovský, and M. Malinovsky, "Influence of NaCl and LiF on the
Aluminium Electrolyte," Electrochimica Acta, vol. 15, pp. 259-269, 1970.
[84] G. Bøe, K. Grjotheim, K. Matiašovský, and P. Fellner, "Electrolytic Deposition of Silicon
and of Silicon Alloys Part II: Decomposition Voltages of Components and Current
Efficiency in the Electrolysis of the Na3AlF6-Al2O3-SiO2 Mixtures," Canadian
Metallurgical Quarterly, vol. 10, pp. 179-183, 1971.
[85] G. Bøe, K. Grjotheim, K. Matiašovský, P. Fellner, "Electrolytic Deposition of Silicon and
of Silicon Alloys Part III: Deposition of Silicon and Aluminum using a Copper Cathode,"
Canadian Metallurgical Quarterly, vol. 10, pp. 281-285, 1971.
[86] G. Bøe, K. Grjotheim, K. Matiašovský, P. Fellner, "Electrolytic Deposition of Silicon and
of Silicon Alloys Part IV: Preparation of Alloys with a High Content of Silicon, and
Silicon Refining," Canadian Metallurgical Quarterly, vol. 11, pp. 463-467, 1972.
[87] K. Grjotheim, K. Matiašovský, and P. Fellner, "Preparation of Silicon Alloys by
Electrolysis of Cryolite- Silica Melts," Revue Roumaine de Chimie, vol. 17, pp. 819-829,
1972.
[88] K. Grjotheim, K. Matiašovský, and P. Fellner, "Some aspects of the electrolytic
preparation of silicon alloys," Chemické Zvesti, vol. 27, pp. 165-171, 1973.
[89] P. Fellner and K. Matiašovský, "Electrolytic Silicide Coating in Fused Salts,"
Electrodeposition and Surface Treatment, vol. 3, pp. 235-244, 1975.
[90] Y. A. Zyazev and A. I. Ezrielev, Siberian Trudy Agronomi Fizik, vol. 13, p. 32, 1966.
[91] R. J. Snow, B. J. Welch, "Reactions in the Cryolite-Silica System," Proceedings -
Australasian Institute of Mining and Metallurgy, vol. 241, pp. 81-86, 1972.
[92] P. K. Gairola, R. K. Tiwari, and A. Ghosh, "Rates of Dissolution of a Vertical Nickel
Cylinder in Liquid Aluminum Under Free Convection," Metallurgical Transactions, vol.
2, pp. 2123-2126, 1971.
Page 86
77
[93] S. Sokhanvaran, S. Thomas, and M. Barati, "Charge Transport Properties of Cryolite–
Silica Melts," Electrochimica Acta, vol. 66, pp. 239-244, 2012.
[94] M. Barati and K. S. Coley, "Electrical and Electronic Conductivity of CaO-SiO2-FeOx
Slags at Various Oxygen Potentials: Part I. Experimental Results," Metallurgical and
Materials Transactions B, vol. 37B, pp. 41-49, 2006.
[95] S. L. Schiefelbein, "A New Technique to Measure the Electrical Properties of Molten
Oxides," in Metallurgical Engineering and Materials Science. vol. PhD: Carnegie Mellon
University, 1989.
[96] Y. Ohta, A. Miyanaga, K. Moringa, and T. Yanagase, "Study on the Measuring of the
Electrical Conductivity of Ionic Solutions and Melts," Journal of the Japan Institute of
Metals, vol. 45, pp. 1036-1043, 1981.
[97] A. Silný and B. Haugsdal, "Electrical conductivity measurements of corrosive liquids at
high temperatures," Review of Scientific Instruments, vol. 64, pp. 532-537, 1993.
[98] R. Tamamushi and K. Takahashi, "Instrumental study of Electrolytic Conductance
Measurements using Four-Electrode Cells," Journal of Electroanalytical Chemistry and
Interfacial Electrochemistry, vol. 50, pp. 1036–1043, 1974.
[99] G. J. Janz, "Thermodynamic and Transport Properties of Molten Salts: Correlation
Equations for Critically Evaluated Density, Surface Tension, Electrical Conductance, and
Viscosity Data," Journal of Physical and Chemical Reference Data, vol. 17, 1988.
[100] T. Magnusson and L. Arnberg, "Density and Solidification Shrinkage of Hypoeutectic
Aluminum-Silicon Alloys," Metallurgical and Materials Transactions A, vol. 32, pp.
2605-2613, 2001.
Page 88
79
Appendix A – Phase Diagrams generated by FactSageTM
Phase Diagram of Silicon – Copper
Page 89
80
Phase Diagram of Silicon – Aluminium
Page 90
81
Phase Diagram of Silicon – Tin
Page 91
82
Phase Diagram of Alumina – Cryolite
Page 92
83
Phase diagram of Lithium Fluoride – Cryolite
Page 93
84
Appendix B – Derivation of Equation 7
Deriving the equation for mass transfer coefficient
General mass transfer equation:
m = mass of silica rod dissolving in melt (g) (only considering the portion inserted in the melt)
t = time of dissolution (h)
k = mass transfer coefficient (cm/h)
A = area of silica rod exposed to the melt (cm2)
CS = saturation concentration of silica in melt at the temperature of experiment (g/cm3)
CB = concentration of silica in the bulk at time t (g/cm3)
Mass of rod inserted in the melt:
r = radius of silica rod
h = length of rod inserted in melt
ρ = density of silica rod
Radius of silica rod inserted in melt in terms of mass:
√
Area of silica rod exposed:
(Area should be A = πr2 + 2πrh but πr
2 is negligible because the silica rod rests on the bottom of
the crucible)
Area in terms of mass:
√
√
√
Bulk concentration:
m0 = initial mass of silica rod inserted in melt
Page 94
85
V = volume of melt
Some numbers that can be assumed constant:
V = volume of melt does not change much with dissolution of silica because maximum amount
of silica dissolved is less than 9g into a melt of 200g. Also at each time (15min, 30min, 1, 2, 4,
8hrs) a rod of initial radius of 1.25cm is inserted and the change of radius is less than 0.06cm. To
make derivation easier V is assumed constant.
h = height of melt after the rod is inserted into melt. Also assumed constant for reasons given
above.
Others constants: m0, ρ, π, CS, and k.
When using MS Excel to optimise equation, k and CS will be solved.
Some constants are defined in terms of a single letter to make equations more legible:
√
Deriving the equation:
√ ([
] )
Defining more constants:
Continuing with derivation:
√ (
)
∫
√ ( )
∫
Using substitution integration:
u = √ u2 = m
2 u du = dm
∫
(
)
Page 95
86
∫
Using integral identity:
∫
|
|
∫
∫
|
|
|
|
Since u = √ , using the experimental values obtained, √ > √ .
Define a = C1V, x = u
√ |
√ √
√ √ |
@t = 0, m = m0
Solving for C
√ |
√ √
√ √ |
Hence:
√
|
√ √
√ √ | |
√ √
√ √ |
Rearranging the equation to get mass in terms of time:
First defining more constants:
|√ √
√ √ |
√
Continuing:
|√ √
√ √ |
√ √
√ √
Page 96
87
√ √ (√ √ )
√ √
√ √
@ t = 0
Since:
|
√ √
√ √ |
|
√ √
√ √ |
√ √
√ √
Also:
√
√
Therefore:
Just looking at:
√
√
√ √
√ √
√
√
√ √
√ √
( √ )(√ √ ) ( √ )(√ √ ) √ √ ( √ )
( √ )(√ √ )
( √ )(√ √ ) ( √ )(√ √ ) √ √ ( √ )
( √ )(√ √ )
Page 97
88
( √ √ √ √ √ √ )
( √ √ √ √ √ √ )
( √ √ √ √ √ √ )
( √ √ √ √ √ √ )
( √ √ √ √ √ √ )
( √ √ √ √ √ √ )
√ √
√ √
√ √
(√ √ )
Returning to original equation:
@ t = ∞
CSV = the total mass of silica dissolved in the melt.
Page 98
89
Appendix C – Effect of considering V and h as constants on k and CS
This section will explain why considering volume of the melt, V, and the height of the melt, h,
does not considerably change the values calculated for k and CS calculated for Equation 7. To
simplify calculations only limiting values will be used:
Maximum measured amount of silica dissolved into CR+LiF melt = 9g
Maximum measured change in radius of silica rod: 0.06cm
Initial mass of melt = 200g
Maximum density measured of CR+LiF with silica addition = 2.5g/cm3
Minimum density measured of CR+LiF without silica = 2.0g/cm3
Calculate initial volume of melt: V = m/ρ = 200/2.0 = 100 cm3
Calculate final volume of melt: V = m/ρ = 209/2.5 = 83.6 cm3
Associated change in height = 0.5cm
The values of k and CS are within experimental error of each other.
Plotting mass transfer coefficient:
0
1
2
3
4
5
6
7
8
9
10
910 930 950 970 990 1010
Mass
Tra
nsf
er C
oef
fici
ent
(cm
/h)
Temperature (°C)
k with volume considered constant
k with volume change
Page 99
90
Plotting solubility limit:
1.7
1.9
2.1
2.3
2.5
2.7
2.9
3.1
3.3
3.5
3.7
910 930 950 970 990 1010
Solu
bil
ity L
imit
(w
t%)
Temperature (°C)
Cs with volume considered constant
Cs with volume change
Page 100
91
Appendix D – XRD analysis of white deposit on silica rod in dissolution behaviour experiment