Solubility of Copper in Silicons

8/11/2019 Solubility of Copper in Silicons

1/231

I

Characterization of the Cu-Si System andUtilization of Metallurgical Techniques in Silicon

Refining for Solar Cell Applications

by

Aleksandar Mitrainovi

A thesis submitted in conformity with the requirementsfor the degree of Doctor of Philosophy

Department of Materials Science and Engineering

University of Toronto

Copyright by Aleksandar Mitrainovi 2010


2/231

II

Characterization of the Cu-Si System and Utilization of

Metallurgical Techniques in Silicon Refining for Solar CellApplications

Aleksandar Mitrainovi

Doctor of Philosophy

Department of Materials Science and Engineering

University of Toronto

2010

Abstract

Two methods for refining metallurgical grade silicon to solar grade silicon have been

investigated. The first method involved the reduction of impurities from metallurgical

grade silicon by high temperature vacuum refining. The concentrations of analyzed

elements were reduced several times. The main steps in the second refining method

include alloying with copper, solidification, grinding and heavy media separation. A

metallographic study of the Si-Cu alloy showed the presence of only two

microconstituents, mainly pure silicon dendrites and the Cu3Si intermetallic. SEM

analysis showed a distinct boundary between the silicon and the Cu3Si phases, with a

large concentration of microcracks along the boundary, which allowed for efficient

separation. After alloying and grinding, a heavy media liquid was used to separate the


3/231

III

light silicon phase from the heavier Cu3Si phase. Cu3Si residues together with the

remaining impurities were found to be located at the surface of the pure silicon particles,

and should be efficiently removed by acid leaching. Thirty elements were analyzed by

the Inductively Coupled Plasma Mass Spectrometry (ICP) chemical analysis technique.

ICP revealed a several times higher impurity level in the Cu3Si intermetallic than in the

pure silicon; furthermore, the amounts of 22 elements in the refined silicon were reduced

below the detection limit where the concentrations of 7 elements were below 1ppmwand

6 elements were below 2ppmw. The results showed that the suggested method is efficient

in removing impurities from metallurgical grade silicon with great potential for furtherdevelopment.


4/231

IV

Content

1 Introduction .........................................................................................11.1 Conversion Efficiency of Silicon Solar Cells...........................................................................51.2 Energy Payback for Silicon Solar Cells....................................................................................71.3 Future of Silicon Solar Cells.......................................................................................................81.4 MGS as Possible Primary Resource in SGS Production .......................................................91.5 Objectives...................................................................................................................................11

2 Silicon Production and Refining .......................................................122.1 MGS Production .......................................................................................................................132.2 SGS Production - Conventional Processes............................................................................15

2.2.1 Siemens Process ...................................................................................172.2.2 Union Carbide Process..........................................................................202.2.3 Ethyl Corporation Process.....................................................................21

2.3 Metallurgical Techniques.........................................................................................................232.3.1 Acid Leaching ......................................................................................232.3.2 Alloying ...............................................................................................232.3.3 Directional Solidification......................................................................242.3.4 Reactive Gas Blowing...........................................................................262.3.5 Slagging ...............................................................................................262.3.6 High Temperature Vacuum Refining ....................................................27

2.4 Refractory Material for Si Refining........................................................................................272.4.1 Mullite..................................................................................................282.4.2 Graphite................................................................................................302.4.3 Glass.....................................................................................................30

2.4.4 Silicon Carbide .....................................................................................313 Impurity Detection in High Purity Silicon.........................................33

3.1 Inductively Coupled Plasma Mass Spectrometry (ICP) .....................................................343.2 Atomic Absorption Spectrophotometry (AAS)...................................................................353.3 Standard Titration Technique (STT)......................................................................................363.4 Glow Discharge Mass Spectrometry (GDMS)....................................................................363.5 LECO - Combustion Techniques...........................................................................................373.6 Recommended Analytical Techniques for Cu-Si Alloys....................................................37

4 Vacuum Refining...............................................................................384.1 Material Balance of Silicon Refining By Vacuum Technique ..........................................40

4.2 Experimental Investigation of Impurity Reduction in Silicon by the Vacuum Technique......................................................................................................................................................43

4.3 Vacuum Technique - Experimental Procedure ....................................................................444.4 Impurity Reduction in Metallurgical Grade Silicon.............................................................504.5 Influence of the Holding Temperature on Impurity Reduction During Vacuum

Treatment....................................................................................................................................514.6 Impurity Reduction During Vacuum Treatment..................................................................544.7 Vacuum Refining Results........................................................................................................56


5/231

V

5 Properties of the Cu-Si System ..........................................................595.1 Copper Diffusivity in Silicon...................................................................................................625.2 Copper Solubility in Silicon.....................................................................................................655.3 Copper Distribution Sites in Silicon .......................................................................................665.4 Impurities in Cu-Si Alloys .......................................................................................................69

6 Metallurgical Refining of MGS ......................................................... 716.1 Chemical Composition of Starting Copper and Silicon......................................................736.2 Apparatus....................................................................................................................................746.3 Alloying......................................................................................................................................766.4 Cooling Rates.............................................................................................................................766.5 Generation of Size Fractions....................................................................................................776.6 Heavy Media Separation..........................................................................................................806.7 Chemical Analysis....................................................................................................................84

7 Metallographic Analysis .................................................................... 857.1 Oxidation Behavior...................................................................................................................85

7.2 Microscopy Findings................................................................................................................867.2.1 Micro-constituents Analysis..................................................................877.2.2 Phase Boundary Condition.................................................................... 897.2.3 Dendrite Thickness Measurements........................................................ 907.2.4 SEM Analysis....................................................................................... 937.2.5 EDX Analysis....................................................................................... 96

7.3 Image Analysis ..........................................................................................................................997.3.1 Area Fraction Determination............................................................... 1007.3.2 Homogeneous Particles....................................................................... 102

7.4 Silicon Dendrite Origin..........................................................................................................1057.4.1 Temperature Regime Influence on Solidification Characteristics......... 106

7.4.2 Silicon Dendrite Formation................................................................. 1078 Characterization of the Cu-Si System..............................................113

8.1 DTA Experimental Procedure ..............................................................................................1138.2 DTA Results ............................................................................................................................1168.3 XRD Analysis..........................................................................................................................129

9 Impurity Contents ............................................................................ 1329.1 Copper Chemical Analysis....................................................................................................1339.2 Impurity Contents....................................................................................................................1419.3 Elemental Concentrations in Light and Heavy Media......................................................1439.4 Silicon Recovery .....................................................................................................................155

9.5 Recovery of Silicon and Driving Forces for HLMS..........................................................1609.6 Copper Removal.....................................................................................................................1629.7 Impurity Removal...................................................................................................................1649.8 Optimal Particle Size ..............................................................................................................1669.9 Distribution Ratio....................................................................................................................1679.10 Maximum Distribution Ratio................................................................................................169


6/231


7/231

VII

Table of Figures

Figure 1.1. Annual demand and availability of silicon feedstock [1-2] ...................... 1Figure 1.2. Price of the silicon based solar module in U.S. [1-3] ...............................2Figure 1.3. Cost breakdown of silicon solar cell components. [1-6]...........................4Figure 1.4. Energy efficiency improvement in silicon solar cells measured under the

global AM1.5 spectrum (1000w/m2) at a cell temperature of 25oC. [1-12]...............................................................................................................5

Figure 1.5. Relationship between purity and the cost of silicon. .............................. 10Figure 2.1. Schematic representation of a furnace for production of metallurgical

grade silicon [2-5]................................................................................. 15Figure 2.2. Schematic representation of the traditional Siemens reactor [2-8] and

positioning of the Si rods inside the reactor [2-9]. .................................18Figure 2.3. A schematic representation of a fluidized bed reactor for polysilicon

production [2-11]..................................................................................22Figure 2.4. Al2O3-SiO2phase diagram where mullite is an intermediate compound

with ideal stoichiometry 3Al2O3-2SiO2. [2-25]......................................29Figure 3.1. An approximate detection limit for the ICP technique used at the

University of Toronto [3-19]................................................................. 34Figure 3.2. Elements determined by the AAS [3-20]............................................... 35Figure 4.1. Vapor pressure of various elements and same oxides. ........................... 38Figure 4.2. Relationship between vapor pressure (Pa) and temperature (K) for

phosphorus and silicon [4-18]. ..............................................................39Figure 4.3. Flow chart of the suggested Vacuum Route. Steps in highlighted fields

were tested in current project. ...............................................................43Figure 4.4. Experimental setup of the vacuum refining trials................................... 44Figure 4.5. Cross section of the furnace used in the vacuum experiments................ 45Figure 4.6. Temperature regime for the vacuum experiments.................................. 46Figure 4.7. Lance position during vacuum experiments........................................... 49Figure 4.8. Impurity path during purging with argon during vacuum refining

experiments. .........................................................................................49Figure 4.9. Mo, Pb and Zn in MGS before and after vacuum treatment (24h).......... 51Figure 4.10. Co, Cr, W and Zr in MGS before and after vacuum treatment (24h)......52Figure 4.11. Mn, Ni, Cu and V in MGS before and after vacuum treatment (24h). .... 52Figure 4.12. Al, Fe and Mg in MGS before and after vacuum treatment (24h). ......... 53

Figure 4.13. Cu reduction from MGS during vacuum experiments after holdingfor 24h.................................................................................................54

Figure 4.14. Zn and Zr reduction from the MGS during vacuum experiments afterholding for 24h. ....................................................................................55

Figure 4.15. Al, Fe and Mg reduction from the MGS during vacuum experiments afterholding for 24h. ....................................................................................55

Figure 5.1. Copper-silicon binary phase diagram [5-5]............................................ 61Figure 5.2. Diffusivity of copper in silicon [5-11]................................................... 63


8/231

VIII

Figure 5.3. Solubility of copper in silicon for various temperatures......................... 65Figure 5.4. Possible copper distribution sites in a silicon sample. 1) Lattice defect, 2)

Grain Boundary, 3) Sample surface, 4) Precipitation into Cu3Si............66Figure 5.5. Characteristic microconstituents distribution in 50wt%Cu-Si alloy

solidified under (a) slow cooling regime and (b)fast cooling regime.

Dark fields are Si dendrites while bright fields are Cu3Si intermetallic.. 67Figure 5.6. Formation of the Cu-silicides around point defect and its growth in bulkSi . Si lattice constant 5.43; Cu lattice constant 3.61 ; Si atomic radius1.46 ; and Cu atomic radius 1.57 . ...................................................68

Figure 5.11. Proposed refining techniques in order to purify metallurgical silicon tosolar grade silicon. ................................................................................70

Figure 6.1. Flow chart of the proposed metallurgical route...................................... 71Figure 6.2. Experimental setup for alloying silicon with copper. ............................. 74Figure 6.3. Reaction tube during alloying. ..............................................................75Figure 6.4. Heating-cooling regime.........................................................................77Figure 6.5. Fraction generation. ..............................................................................78

Figure 6.6. Sample collection: A view of the specimen after removing from thefurnace (upper figure), collected fractions from section B after crushingand screening (bottom-left) and middle cross segment from section C(bottom-right). ......................................................................................79

Figure 6.7. Three distinctive layers formed after heavy media separation................ 80Figure 6.8. Heavy media separation procedure........................................................83Figure 7.1. Weight gain of fine grinded Si-50wt%Cu alloy (size less than 38m) in

laboratory air at room temperature. .......................................................86Figure 7.2. Secondary electron images with the two microconstituents phases present

in 50Cu-Si hypereutectic alloy. a) close to the wall, b) region between theedge and the center of the sample and c) central part of the sample. ...... 87

Figure 7.3. Secondary electron image with a number of loose fractions at the edge ofthe Si and CuxSi phases.........................................................................88Figure 7.4. Secondary electron image with a number of microcracks along the

microconstituents boundary in the Si-50%wtCu alloy; carbon coated,20kV [7-3]............................................................................................89

Figure 7.5. The middle cross section of the as is solidified specimen. ..................... 91Figure 7.6. Average dendrite thickness calculation procedure. I ) part of the sample,

II ) straitening the circle and III ) counting Si and thickness measurement..............................................................................................................92

Figure 7.7. Loose Si particles after gravity separation mounted onto the SEM samplestand [7-3]. ...........................................................................................93

Figure 7.8. Loose silicon rich particle after gravity separation with CuxSi residues atthe surface [7-3]....................................................................................94Figure 7.9. Loose silicon rich particles after gravity separation with CuxSi residues at

the surface with corresponding EDX analyses in Figure 7.10 [7-3]........95Figure 7.10. EDX analysis of the loose silicon rich particles after gravity separation

with CuxSi residues at the surface. ........................................................95Figure 7.11. Si-50wt%Cu alloy fraction prior to crushing and gravity separation

prepared for EDX analysis, mounted in epoxy. .....................................96


9/231

IX

Figure 7.12. EDX analysis of the Si-50wt%Cu alloy fraction prior to gravityseparation. Upper diagram is related to point 1 in Figure 7.11 and bottomdiagram is related to point 2 in Figure 7.11. ..........................................97

Figure 7.13. SEM figures of different size silicon-copper alloy particles before heavymedia separation. a) Particle size is 106-500m; b) particle size is 500-

1000m, and c) particle size is above 1000m. .....................................99Figure 7.14. Mixed particle used to determine total area, Si area and Cu3Si area.....100Figure 7.15. Homogeneous and mixed particles in the 50Cu-Si alloy...................... 103Figure 7.16. Area fractions of homogeneous particles and Si particles with up to 1%

CuxSi prior to gravity separation. ........................................................ 104Figure 7.17. Temperature regime during solidification............................................ 107Figure 7.18. Solidification of the Cu-Si system. a) Forces present in Cu-Si system and

heat flow, b) mass flow, c-e) Si dendrites formation, growth and path ofthe Si particles. ................................................................................... 108

Figure 7.19. A view of the sample after removal from the furnace, method by whichthe sample was sectioned, up-right surface prepared for metallographic

analysis...............................................................................................109Figure 7.20. Morphology of the Cu-Si alloy sample with three regions (i, ii and iii)and four layers (1, 2, 3 and 4).............................................................. 110

Figure 7.21. Angle and length of the Si dendrites throughout the 50Cu-Si samplemeasured from the bottom................................................................... 111

Figure 7.22. Angle and length of Si dendrites throughout the 70Cu-Si samplemeasured from the bottom................................................................... 112

Figure 8.1. Selected chemical compositions for 4 alloys generated for DTA analysis............................................................................................................115

Figure 8.2. Characteristic DTA curves for pure silicon and pure copper................ 117Figure 8.3. DTA curves as a function of temperature. ........................................... 118

Figure 8.4. DTA curves for the solidus region for 4 Cu-Si alloys. ......................... 119Figure 8.5. DTA curves for the 50wt%Cu-Si and 70wt%Cu-Si alloys................... 120Figure 8.6. DTA curves for the 83wt%Cu-Si and 87wt%Cu-Si alloys................... 121Figure 8.7. DTA curves with respect to temperature for samples generated from the

light and heavy media. ........................................................................ 124Figure 8.8. DTA curves for solar grade silicon and light media fraction samples... 125Figure 8.9. DTA curves for a 50Cu alloy and a mixture of light and heavy media

fraction samples.................................................................................. 126Figure 8.10. DTA curves for 83Cu alloy and heavy media fraction samples............ 127Figure 8.11. Shape of the DTA curves for a) transformation in solid, b) solidus and c-

d) liquidus. 50wt%Cu, 70wt%Cu and mixture of light and heavy media

samples...............................................................................................128Figure 8.12. XRD of the 50wt%Cu-Si light and heavy media particles. .................. 129Figure 8.13. XRD of the non-oxidized samples with 50wt%Cu, 70wt%Cu and

83wt%Cu chemical composition alloys. .............................................. 130Figure 8.14. XRD of the light and heavy media sample a with possible SiO2peaks.131Figure 9.1. Recommended chemical analysis methods for Cu-Si alloys samples. ICP

- Inductively Coupled Plasma, AAS - Atomic Absorption Spectrometryand STT - Standard Titration Technique ............................................. 132


10/231

X

Figure 9.2. Cu content in the light fraction after gravity separation for the 50Cu-Sialloy (Specimens #1-#4) and average Cu content with standard deviationbars.....................................................................................................134

Figure 9.3. Amount of Cu in the light fraction after gravity separation for the 50Cu-Si and 70Cu-Si alloys.......................................................................... 135

Figure 9.4. Amount of Cu in the heavy fraction after gravity separation for the 50Cu-Si alloy (Specimens #1-#4). ................................................................ 139Figure 9.5. Amount of Cu in the heavy fraction after gravity separation for the 70Cu-

Si alloy. .............................................................................................. 140Figure 9.7. Concentrations of selected impurities after gravity separation of the Si-

50Cu alloy. ......................................................................................... 141Figure 9.8. Concentrations of Al, Ca, Cr, Fe and Zn after gravity separation in the Si-

50Cu alloy. ......................................................................................... 142Figure 9.10. Concentration of Al in the light and heavy media................................ 144Figure 9.11. Concentration of Ag in the light and heavy media............................... 144Figure 9.12. Concentration of Ba in the light and heavy media. .............................. 145

Figure 9.13. Concentration of Bi in the light and heavy media. ............................... 145Figure 9.14. Concentration of Ca in the light and heavy media. .............................. 146Figure 9.15. Concentration of Cd in the light and heavy media. .............................. 146Figure 9.16. Concentration of Co in the light and heavy media. .............................. 147Figure 9.17. Concentration of Cr in the light and heavy media................................ 147Figure 9.18. Concentration of La in the light and heavy media................................ 148Figure 9.19. Concentration of Fe in the light and heavy media................................ 148Figure 9.20. Concentration of Mn in the light and heavy media. ............................. 149Figure 9.21. Concentration of Mo in the light and heavy media. ............................. 149Figure 9.22. Concentration of Ni in the light and heavy media................................ 150Figure 9.23. Concentration of Pb in the light and heavy media................................ 150

Figure 9.24. Concentration of Sc in the light and heavy media................................ 151Figure 9.25. Concentration of Sr in the light and heavy media. ............................... 151Figure 9.26. Concentration of Ti in the light and heavy media. ............................... 152Figure 9.27. Concentration of Tl in the light and heavy media. ............................... 152Figure 9.28. Concentration of Zn in the light and heavy media. .............................. 153Figure 9.29. Concentration of Zr in the light and heavy media. ............................... 153Figure 9.30. Concentration of V in the light and heavy media................................. 154Figure 9.31. Concentration of W in the light and heavy media. ............................... 154Figure 9.32. Fashion which samples were prepared and weighed for recovery analysis.

...........................................................................................................155Figure 9.33. Apparent silicon recovery in the 50Cu-Si alloy for 4 specimens. ......... 156

Figure 9.34. Silicon recovery from the 50Cu-Si alloy for 4 specimens. ................... 158Figure 9.35. Forces present in the system during gravity separation. ....................... 160Figure 9.36. Cu removal from the 50Cu-Si alloy..................................................... 163Figure 9.37. Removal of the impurities after gravity separation in the 50Cu-Si. .......... 164Figure 9.38. Removal of Al, Ba, Cr, Pb, Sr and Zn after gravity separation in the

50Cu-Si alloy...................................................................................... 165Figure 9.39. Removal of Cu for different particle size derived from the ICP chemical

analyzes..............................................................................................166


11/231

XI

Figure 9.40. Distribution ratio of impurities in heavy over light fraction in the 50Cu-Sialloy after gravity separation [7-3]. ..................................................... 167

Figure 9.41 Distribution ratio for Al, Ca, Cr, Pb and Zn in heavy over light fraction inthe 50Cu-Si alloy after gravity separation [7-3]...................................168

Figure 9.42. Maximum distribution ratio of impurities in 50Cu-Si alloy after gravity separation.

...........................................................................................................170Figure 9.43. Maximum distribution ratio of impurities in 75Cu-Si alloy after gravity separation.

...........................................................................................................170Figure 11.1. Suggested experimental setup for vacuum solidification investigation

using modified PoDFA unit. ............................................................... 178Figure A1. The equipment used during mixing and melting of metallurgical silicon

and copper: 1. Argon gas; 2. Mass flow controller; 3. Copper getterfurnace; 4. Lance; 5. Reaction tube; 6. Electric furnace; 7. Temperaturecontroller. ........................................................................................... 192

Figure A2. Equipment used for screening Cu-Si alloys. Tyler Canada, FH Brass Pantype screens. ....................................................................................... 192

Figure A3. Equipment used for gravity separation of the Cu-Si alloys. .................. 192Figure A4. Analytical balance used for weighting samples. ................................... 192Figure A5. Furnace chamber temperature measured with controller thermocouple

sited at the edge of the crucible and R-type thermocouple placed in themiddle of the crucible. ........................................................................ 193

Figure A6. Time required for furnace chamber temperature (measured with R-typethermocouple placed in the middle of the crucible) to reach controllergiven temperature (measured at the edge of the crucible). ................... 193

Figure A7. Vapor Pressure of Elements ................................................................ 194Figure A8. Standard Free Energies of Formation, Ellingham Diagram................. 196Figure A9. Standard Free Energies of Formation of chlorides............................... 197

Figure A10. DTA curves for non-oxidized samples. ............................................... 199Figure A11. DTA curve for non-oxidized solar grade sample. ................................ 199Figure A12. DTA curve for non-oxidized 50wt%Cu-Si sample. ............................. 200Figure A13. DTA curve for non-oxidized 70wt%Cu-Si sample. ............................. 200Figure A14. DTA curve for non-oxidized 83wt%Cu-Si sample. ............................. 201Figure A15. DTA curve for non-oxidized 87wt%Cu-Si sample. ............................. 201Figure A16. DTA curve for non-oxidized electronic copper sample........................ 202Figure A17. DTA curve for light media sample. ..................................................... 202Figure A18. DTA curve for heavy media sample.................................................... 203Figure A19. DTA curves for light and heavy media samples. ................................. 203Figure A20. Shift intensities for Si, Cu and Cu3Si phases. ...................................... 204

Figure A21. Shift intensities for SiO2. .................................................................... 205Figure A22. SEM figure for 38-106m particle size before heavy media separation for50wt%Cu-Si alloy............................................................................... 206

Figure A23. Counting the number of particles in 38-106m particle size sample beforeheavy media separation (50wt%Cu-Si alloy)....................................... 206

Figure A24. SEM figure for the 106-500m particle size before heavy mediaseparation for 50wt%Cu-Si alloy. ....................................................... 207


12/231

XII

Figure A25. Count of the particles in 106-500m particle size sample before heavymedia separation (50wt%Cu-Si alloy). ................................................ 207

Figure A26. SEM figure for the 1000+m particle size before heavy media separationfor 50wt%Cu-Si alloy. ........................................................................ 208

Figure A27. Count of the particles in 1000+m particle size sample before heavy

media separation (50wt%Cu-Si alloy). ................................................ 208


13/231

XIII

List of Tables

Table 2.1. Segregation coefficients for various impurities in silicon [2-21]............ 25Table 2.2. Typical physical and mechanical properties of some refractory material.

.............................................................................................................28Table 2.3. Typical physical and mechanical properties of the mullite..................... 28Table 2.4. Typical physical and mechanical properties of graphite [2-26]. .............30Table 2.5. Typical physical and mechanical properties of the carbide. ...................31Table 3.1. Most suitable measurement techniques for trace amounts of impurities in

pure silicon [3-18].................................................................................33Table 4.1: Parameters for impurity reduction c from liquid silicon......................... 40Table 4.2. Experimental design for the Vacuum Route procedure.......................... 47

Table 4.3. Chemical composition (ppmw) of the metallurgical grade silicon (MGS)used in the experiments.........................................................................48Table 4.4. Impurity concentration in ppmwin metallurgical grade silicon before and

after vacuum refining. ...........................................................................56Table 4.5. Impurity Reduction Factor after vacuum refining.................................. 57Table 5.1. Atomic structure and physical properties of the copper and silicon........59Table 5.2. Copper diffusivity in silicon.................................................................. 62Table 6.1. Chemical composition (ppmw) of the electrolytic copper (EC) and

metallurgical grade silicon (MGS) used in experiments.........................73Table 6.2. Different size fractions used in the experiments. ................................... 78Table 7.1. Weight gain of Si-50wt%Cu alloy fraction with size less than 38m.....85

Table 7.2. Local thickness of the dendrites in the middle cross-section of thespecimen...............................................................................................92Table 7.3. EDX analysis of mixed particle close to the phase boundary with

respective standard deviation for silicon in the bright field....................98Table 7.4. Number of particles and total analyzed area for samples prior to

separation. .......................................................................................... 100Table 7.5. Area fraction of silicon prior to and after gravity separation for 50Cu-Si

alloy*..................................................................................................101Table 7.6. Number of homogeneous particles in 50Cu-Si alloy............................ 102Table 7.7. Characteristic parameters for hypereutectic Cu-Si alloys calculated by

FactSage Chemical Thermodynamics Application. ............................. 106

Table 8.1. Weight of the DTA samples used in experiments. ............................... 114Table 8.2. Characteristic temperatures during heating for non-oxidized samples.. 116Table 8.3. Characteristic temperatures during cooling for non-oxidized samples. 116Table 8.4: Weight and approximate chemical composition of the light and heavy

media samples. ................................................................................... 122Table 8.5. Characteristic temperatures during heating for light and heavy media

samples...............................................................................................122


14/231

XIV

Table 8.6. Characteristic temperatures during cooling for light and heavy mediasamples...............................................................................................123

Table 9.1. ICP and AAS results for light media sample, 50wt%Cu-Si.................. 136Table 9.2. Average copper concentration in different fractions for 50wt%Cu-Si alloy

with standard deviation**: .................................................................. 137

Table 9.3. Polynomial coefficients of the concentration curve in the light media forthe 50Cu-Si and 70Cu-Si alloys. ......................................................... 138Table 9.4. Weights of the light and heavy fractions after heavy media separation:159Table 11.1. Tolerable impurity concentration before directional solidification and

achieved concentration in current project (ppmw)................................176


15/231

1

1 Introduction

Our planet receives approximately 1.21017W of solar power, while the rate of current

worldwide energy consumption is approximately 10,000 times smaller at around

1.31013W [1-1]. This indicates that the earth receives more solar energy in an hour

than the total energy consumed by humans in an entire year. Solar energy alone has the

potential capacity to meet the planets entire energy needs in the future. Additionally, an

increase in global energy consumption, higher energy prices and public awareness of

global warming has unlocked the solar cell market. In 2003, the photovoltaic (PV)

industry had a growth rate of more than 30%, a situation that many other industries can

only dream about [1-2]. Meanwhile, the most recent analysis of the silicon solar cell

market [1-3] shows a growth rate of 110% in 2008 in installations and with $37.1 billion

in global revenues. The majority of solar cells are made of silicon, and experts believe

that it will take at least a decade before other photovoltaic technologies based on other

materials could be competitive.

Figure 1.1. Annual demand and availability of silicon feedstock [1-2]


16/231

2

The remarkable growth of the photovoltaic industry has caused a shortage of Solar

Grade Silicon (SGS), or silicon with the required chemical purity for photovoltaic

applications, (Fig. 1.1). The primary source of solar grade silicon has been silicon

rejected from the electronic industry. Presently, the production cost for solar grade

silicon, via the labor and energy intensive chemical vapor deposition technique, is a

limiting factor for environmentally friendly solar energy to become one of the major

energy sources (Fig 1.2).

Figure 1.2. Price of the silicon based solar module in U.S. [1-3]

Most solar cells and modules are made of either single crystal or polycrystalline silicon,

which are converted into large wafers of purified silicon. The other major costsinvolved in photovoltaic components are modules, panels, arrays, mounting equipment,

wiring, inverters, grid connection equipment, service and labor to assemble,

management and administration costs and costs of permits etc. A summarized cost

breakdown of a silicon solar cell is given in Figure 1.3, where it can be seen that the


17/231

3

solar grade silicon contributes to about 50% of the cost of the solar cell module (Figure

1.3,a). Furthermore, module cost contributes into the final system setup with about 55%

(Figure 1.3,b) bringing the cost of the silicon in the final cost to 25-30%. In addition,

looking at the energy required for different parts production and system installation,

shows that silicon contributes far more than any other component, about 59% (Figure

1.3,c). This leads to the conclusion that the greatest avenue for price per Watt

reduction is in the silicon refining process where price per watt is a measure of the cost

of the solar panel expressed as a ratio between the price of the panel and the nominal

number of watts. Implementing new production and refining techniques for solar grade

silicon could reduce the price of a silicon module to $3 per Watt, which some

economists denote as a critical barrier for the transition of the solar cell industry from

small contractor businesses into larger companies. Involvement of larger companieswill result in significant cost reduction in other major aspects involved in photovoltaic

systems that can initiate a further cost reduction to $1 per Watt [1-4]. Reaching this

benchmark will be the turning point from which markets will emerge and grow without

any government aid and that could be the transition point for the solar cell industry from

the alternative source to a reliable, economical source of energy [1-5]. The greatest

limitations for silicon based photovoltaic systems are the efficiency and price of the

silicon solar cells.


18/231

4

Figure 1.3. Cost breakdown of silicon solar cell components. [1-6]

a)

b)

c)


19/231

5

1.1 Conversion Efficiency of Silicon Solar Cells

Henri Becquerel discovered in 1836 the photovoltaic effect or the capacity of some

material systems to generate a voltage when illuminated. Pearson, Chapman, and Fullerfrom the Bell Laboratories, first reported the modern silicon solar cell in 1954

demonstrating a 5 to 6% efficiency [1-7]. In the 1960s, solar cells were successfully

developed primarily for powering satellites. Such cells were too expensive for almost

all other applications. Considerable technical progress and cost reductions were gained

from government-funded R&D programs during the energy crisis of the 1970s.

Nowadays, the efficiency of silicon solar cells in laboratory conditions is approaching

the theoretical limit of 29%. Figure 1.4 shows the rise in efficiency of laboratory

produced solar cell's from the 1950s up today.

Figure 1.4. Energy efficiency improvement in silicon solar cells measured under

the global AM1.5 spectrum (1000w/m2) at a cell temperature of 25oC. [1-12]


20/231

6

Basically, the different semiconductor materials or combinations are suited only for

specific spectral ranges. Particularly, silicon transforms 1.1eV of photons

electromagnetic energy into electrical energy. Therefore a specific portion of the radiant

energy cannot be used, because some photons do not have enough energy to activate the

charge carriers, e.g photons below 1.1eV in silicon based solar cells. On the other hand,

surplus photon energy is transformed into heat rather than into electrical energy. In

addition to that, there are optical losses, reflections of incoming rays, electrical

resistance losses, disrupting influences of material contamination, surface effects and

crystal defects, etc. Single loss mechanisms cannot be further improved because of

inherent physical limits imposed by the materials themselves. This leads to a theoretical

maximum level of efficiency, i.e. approximately 29% for crystal silicon.

Research results have demonstrated efficiencies close to 25%, while industrially

produced solar cells have efficiencies ranging from 12% to 18%. Closing the gap

between laboratory and factory efficiencies depends on materials-processing

innovations, composition and purity of the silicon, thermodynamics, phase

transformation, kinetics, heat and mass transfer, hydrodynamics, solid-state physics,

thin-film technology, process control, etc.

To provide some sense of the capabilities of silicon based solar cells, the required area

of solar cells needed to meet the average consumption of an U.S. residence has been

calculated. Under noontime sunshine on a clear summer day, a typical (12% efficient)

silicon solar-cell module with an area of 1m2 would produce about 100W of direct-

current power. On average over the year, this output could be sustained for about six

hours per day, for a daily yield of about 600Wh of energy per square meter of installed

solar cells. The typical U.S. household consumes about 25kWh of electricity per day,

and solar cells comprising about 42m2would be required to provide sufficient power.


21/231

7

1.2 Energy Payback for Silicon Solar Cells

In Figure 1.3, it can be seen that the major cost in the production of a solar module is

producing the solar grade silicon (SGS). Furthermore, purifying and crystallizing thesilicon is the most energy consuming part of the solar cell manufacturing process.

Advantages of producing electricity with photovoltaic devices are zero emissions and no

fossil fuel consumption. These are great environmental benefits, however the

production of pure silicon also consumes energy. The period required for a solar cell to

produce the same amount of energy that was used to create it, is called the Payback

Time. Due to the many factors influencing quality and lifetime of solar cells, the

economy of the solar cells is usually expressed in the Payback Time rather than in the

cost of the solar cell.

An analysis of the Payback Time for solar grade silicon that had originally been rejected

from electronic grade silicon consumers was performed by Erik Alsema [1-13]. His best

estimates of energy used to make frameless PV were 600 kWh/m2 for single crystal

silicon modules and 420 kWh/m2 for polycrystalline silicon. Assuming a 12%

conversion efficiency (standard conditions) and 1700 kWh/m2 per year of available

sunlight energy (the U.S. average is 1800), the calculated payback time is about fouryears for contemporary polycrystalline silicon photovoltaic systems. Projecting

improvements 10 years into the future, where the author assumes a constant supply of

solar grade silicon feedstock by conventional processes and 14% efficiency, energy

payback is reduced to about two years. Replacing the conventional Siemens process

with a direct metallurgical route can reduce payback time to one year. Furthermore,

silicon treatment by a direct metallurgical route for production of SGS can be five times

more energy efficient than the conventional/commercial processes that use more than

200kWh for 1kg silicon produced [1-14].

The U.S. Department of Energy Photovoltaics Program [1-15] estimates that an average

U.S. household producing 1000 kWh of electricity with solar power reduces emissions

by nearly 8 pounds of sulfur dioxide, 5 pounds of nitrogen oxides, and more than 1,400


22/231

8

pounds of carbon dioxide. During its projected 28 years of clean energy production, a

rooftop system with 2-year payback and meeting half of a households electricity use

would avoid conventional electrical plant emissions of more than half a ton of sulfur

dioxide, one-third a ton of nitrogen oxides, and 100 tons of carbon dioxide, giving great

environmental benefits.

1.3 Future of Silicon Solar Cells

For some time now, due to the limited theoretical efficiency, it has been thought that

monocrystalline or polycrystalline silicon solar cells would be made obsolete by the

eventual introduction of very inexpensive alternative solar cells based on thin films of

semiconductors such as cadmium telluride (CdTe), copper-indium diselenide (CIS),

amorphous silicon (-silicon) or organic photovoltaic material. This may happen, but so

far on the basis of cost, reliability, performance, and lifetime, none of these alternatives

appear to be in a position to challenge silicon solar cells.

Indeed, recent reports about silicon properties and photovoltaic capacities at the nano

level seem to have positioned silicon at the forefront as a major future component for

photovoltaic applications. A typical solar cell generates one electron per photon of

incoming sunlight. Conversely, some materials are capable of producing multiple

electrons per photon, but for the first time this effect have been seen in silicon at the

National Renewable Energy Laboratory (NREL), in Golden, CO. Results have shown

that silicon nanocrystals can produce two or three electrons per photon of high-energy

sunlight [1-16]. In most silicon solar cells, the extra energy in blue and ultraviolet light

is wasted as heat. The small size of nanoscale crystals, called quantum dots [1-17, 1-

18], leads to novel quantum-mechanical effects that convert this energy into electrons.

The effect can lead to a new type of solar cell that utilizs the same feedstock as


23/231

9

commercial solar cells but that are more than twice as efficient as today's typical silicon

based photovoltaic. Arthur Nozik [1-19] claims that by generating multiple electrons

from high-energy photons, solar cells made of silicon nanocrystals could theoretically

convert more than 40 percent of the energy in light into electrical power. Furthermore,

concentrating sunlight with mirrors or lenses could raise that figure to well over 60

percent. New technology not only gains from the advantages of efficiency inherent in

producing different numbers of electrons based on the amount of energy in the striking

photon, but a manufacturer does not incur the expensive pure-silicon-crystal-growth

processes. Such situation can make silicon produced by affordable, high capacity,

entirely metallurgical refined a perfect product.

1.4 MGS as Possible Primary Resource in SGS Production

Due to the strong demand for silicon photovoltaic applications, manufacturers of silicon

solar cells and producers of metallurgical grade silicon have the same interest. For

metallurgical grade silicon producers, the increase and inconsistency in energy prices are

a challenge and it is necessary to find products with higher value and stable demand for

which the energy input contributes less to the fixed costs. At the same time, solar cell

manufacturers are looking for reliable, high capacity and economically viable sources of

solar grade silicon. Under the given conditions, it appears that metallurgical grade

silicon is a prefect starting material for silicon solar cells (Figure 1.5). However, with

the demand for extremely high purity products and the necessity to develop complete

new processes, the introduction of this new product into the market involves a great deal

of research and the development of new silicon treatment techniques.


24/231

10

AGSAGSAGSAGS-Alloy Grade SiMGSMGSMGSMGS-Metallurgical Grade SiEGSEGSEGSEGS -Electronic Grade Si

Figure 1.5. Relationship between purity and the cost of silicon.


25/231

11

1.5 Objectives

The aim of the current project is to investigate a set of the low cost metallurgical

techniques for refining Metallurgical Grade Silicon (MGS) to Solar Grade Silicon(SGS), consequently reducing the cost for solar cell production.

With the purpose of advancing the knowledge of silicon refining, particularly in alloying

with copper, separation from solvent metal and surface treatment with the possibility of

creating a new set of process steps, the overall objective was to develop new silicon

refining processes, including:

Determine optimal process parameters for high temperature vacuum treatment

and a copper impurity collection method.

Determine effect of vacuum treatment on impurities removal from the

metallurgical grade silicon.

Determine the feasibility of the newly developed copper impurity collection

method to reduce the amount of impurities in the metallurgical grade silicon.

Determine the mechanisms of the silicon dendrite formation and optimal

solidification conditions for a Cu-Si alloy in order to form clean and large silicon

dendrites.

Determine the recovery of refined silicon after separation.

Determine the copper content in the silicon after refining process.

Remove all impurity elements to solar grade requirements.

Investigate the industrial potential of purifying silicon by copper alloying and

heavy media separation in order to achieve solar grade requirements.


26/231

12

2 Silicon Production and Refining

Silicon is a metalloid, one of only a very few elements that have characteristics of both

metals and non-metals. The melting point of silicon is 1414C (2574F) and the boiling

point is 2355C (4270F). Its density is 2.33 grams per cubic centimeter, and it forms a

face-centered cubic structure with a lattice spacing of 5.43 [5-1]. Silicon is a relatively

inactive element and does not combine with oxygen or most other elements at room

temperature while at higher temperatures, silicon becomes much more reactive. Also,

water, steam, and most acids have very little effect on silicon. Information on the health

effects of silicon is limited. Overall, silicon probably has no positive or negative effects

on human health unless silicon is ground up into a very fine powder that can cause

silicosis [5-2, 5-3].

Silicon, like carbon and other group IV elements, form a face-centered diamond cubic

crystal structure. Silicon, in particular, forms a face-centered cubic structure with a

lattice spacing of 0.5430710nm.

Silicon ApplicationsSilicon, in the form of SiO2, is vital to the construction industry as a principal constituent

of natural stone, glass, concrete and cement. Silicon's greatest impact on the modern

world's economy and lifestyle has resulted from silicon wafers used in the manufacture

of electronic devices such as power transistors, and integrated circuits such as computer

chips.

Silicon Alloys

The largest application of metallurgical grade silicon, representing more than 50% of the

world consumption, is in the manufacture of aluminium-silicon alloys to produce cast

parts, mainly for the automotive industry. Silicon is an important constituent of steel,

modifying its resistivity and ferromagnetic properties. Silicon is added to molten cast


27/231

13

iron as ferrosilicon to improve its performance in casting thin sections, and to prevent

the formation of cementite at the surface.

2.1 MGS Production

Silicon is the second most abundant element in the Earth's crust (about 28%), exceeded

only by oxygen. Many rocks and minerals contain silicon. Examples include sand,

quartz, clays, flint, amethyst, opal, mica, feldspar, garnet, tourmaline, asbestos, talc,

zircon, emerald, and aquamarine. Silicon never occurs as a free element but as silicondioxide, known as silicates. Silicon is commercially produced from its minerals by the

reaction of high-purity silica with coal in an electric arc furnace using carbon electrodes.

This process is named Carbothermic Reduction of Silica and the silicon produced via

this process is called Metallurgical Grade Silicon (MGS). Metallurgical grade silicon

has a purity of at least 95%. Reduction of silicon is accomplished by the carbothermic

reduction of silica where the main chemical reaction is:

SiO2(s) + 2 C (s) = Si (l) + 2 CO (g) (2.1)

Source of the carbon is coal together with woodchips and coke. Quartz and carbon are

carefully selected in order to maximize furnace performance and to reduce emissions of

the SO2and NOXgases. The mixture of the raw material is heated by an intense electric

arc located between the tips of three submerged electrodes and the electrical ground of

the furnace. The quality of the carbon electrodes is an important factor of the MGS

production. A prepared mixture of the raw materials is charged from the top of the

furnace and liquid silicon is collected from the bottom side. The temperature in the

upper zone is below 1900C and the main reactions are:

SiO2(s) + 2 C (s) = Si (l) + 2 CO (g) (2.1)


28/231

14

SiO (g) + 2 C (s) = SiC (s) + CO (g) (2.2)

2 SiO (g) = Si (l) + SiO2(s) (2.3)

Silicon is produced in the inner zone where the temperature is between 1900 to 2100C.

The main reactions in inner zone are:

2 SiO2(l) + SiC (s) = 3 SiO (g) + CO (g) (2.4)

SiO (g) + SiC (s) = 2 Si (l) + CO (g) (2.5)

A schematic description of the furnace is given in Figure 2.1. Raw materials are fed in

small batches with frequent intervals and are distributed on top of the charge. Liquid

silicon is at frequent intervals, drained out from the bottom of the furnace. The silica

fumes, which consist mainly of very fine particles of amorphous silica less than 1 m,

are passed through filter cloths installed in large bag-house systems adjacent to the

furnaces. The collected amorphous silica finds applications as additives in concrete and

refractories. Depending on the quality of the raw materials used and the operationalstrategy, the silicon yield as metallurgical silicon ranges from 80 to 90%, the balance

resulting in silica fume [2-4].

The liquid crude silicon contains normally less than 3% impurities depending on the raw

materials and the type of electrodes. The main impurities are Fe: 0.21%, Al: 0.40.7%,

Ca: 0.20.6%, Ti: 0.10.02%, C: 0.10.15%


29/231

15

Figure 2.1. Schematic representation of a furnace for production of metallurgical

grade silicon [2-5].

2.2 SGS Production - Conventional Processes

Impurities in the ppb(a)range are required for the electronic grade silicon supplied to the

semiconductor industry. The ultra-high purity is needed to ensure rigorous

semiconductor properties in the silicon crystals.

Many processes to produce polysilicon have been tested, patented and a few operated formany years. Only three large commercial processes are currently active:

1. The most popular process is based on the thermal decomposition of trichlorosilane

(SiHCl3) at 1100C on a heated silicon rod placed inside a deposition chamber.


30/231

16

This process, which was developed in the late fifties, is commonly referred to as

the Siemens process with reference to the company that carried out its early

development.

2. In a more recent process developed by Union Carbide Chemicals in the United

States of America, the trichlorosilane has been replaced by monosilane (SiH4), but

the principle of decomposition on a heated silicon rod inside a closed deposition

chamber is maintained. This process, presently run by the company Advanced

Silicon Materials, LLC. The Union Carbide process has gained a significant

market acceptance in the past 15 years.

3. Finally, in the third process, also making use of monosilane SiH4, the heatedsilicon rod in the closed reaction chamber has been replaced by a heated fluidized

bed containing silicon particles. The particles act as seeds on which SiH4 is

continuously decomposed to larger granules of hyper-pure silicon. Unlike

Siemens or Union Carbide, this process is a continuous one. This process is

known as the Ethyl Corporation process, after the name of the US chemical

company that developed it and presently is run by the US corporation MEMC in

Pasadena, Texas.

The respective features, advantages and disadvantages of these different routes are

described in the following sections.


31/231

17

2.2.1 Siemens Process

Trichlorosilane (HSiCl3) is prepared by hydrochlorination of metallurgical grade silicon

in a fluidised bed reactor [2-6, 2-7, 2-8]:

Si (s) + 3 HCl = HSiCl3+ H2 (2.6)

This reaction occurs at 350C normally without a catalyst. A competing reaction is

contributing to the formation of unsuitable tetrachlorosilane in molar proportion of 10 to

20%:

Si (s) + 4 HCl = SiCl4+ 2 H2 (2.7)

Trichlorosilane is chosen because of its high deposition rate, its low boiling point

(31.8C) and its comparatively high volatility and hence the ease of purification with

respect to boron and phosphorus down to the ppb level. The suitable trichlorosilane

undergoes a double purification through fractional distillation, the first step removing

the heaviest components resulting from the direct synthesis and the second step

eliminating the components lighter than trichlorosilane, also called volatiles. High-purity trichlorosilane is then vaporised, diluted with high-purity hydrogen and

introduced into the deposition reactors. The gas is decomposed onto the surface of

heated silicon seed rods, electrically heated to about 1100C, and growing large rods of

hyperpure silicon. The main reactions are:

H2+ HSiCl3= Si + 3 HCl (2.8)

2 SiHCl3= SiH2Cl2+ SiCl4 (2.9)

SiH2Cl2= Si + 2 HCl (2.10)

HCl + HSiCl3= SiCl4+ H2 (2.11)

The stream of reaction by-products, which leaves the reactor, contains H2, HCl, SiHCl3,

SiCl4and SiH2Cl2.


32/231

18

A schematic representation of the Siemens reactor is given in Figure 2.2. The Siemens

process is highly energy consuming, a major part of the energy being dispersed and lost.

To avoid deposition on the inner surfaces of the reaction chamber, they have to be

cooled. The decomposition chamber consists of a steel bell jar where 30 or more

inverted U-rods are placed in each reactor [2-7].

Figure 2.2. Schematic representation of the traditional Siemens reactor [2-8] and

positioning of the Si rods inside the reactor [2-9].

As reactions and equilibriums (2.8) to (2.11) show, the deposition process generates by-

products. Unfortunately, for each mole of Si converted to polysilicon, 3 to 4 moles are

converted to SiCl4, binding large amounts of chlorine and valuable silicon. With

polysilicon production growing much faster than silicones and pyrogenic silica, the

question is whether to eliminate or recycle the tetrachlorosilane. Recycling the by-

product back to the valuable starting material is generally a preferred solution. There are

two basic chemical processes applicable to reconvert SiCl4to SiHCl3:


33/231

19

1. The high temperature reduction of silicon tetrachloride with hydrogen:

SiCl4+ H2= SiHCl3+ HCl (2.12)

At about 1000C, a 1:1 molar mixture of SiCl4and H2produces approximately

20 to 25% molar SiHCl3 in the gaseous mixture. This process requires a fair

amount of electrical energy but has a distinct advantage that the trichlorosilane

produced is of very high quality because both reactants, silicon tetrachloride and

hydrogen, are basically electronic grade when produced by equations (2.8) and

(2.11).

2. The hydrogenation of silicon tetrachloride in a mass bed of metallurgicalsilicon:

3SiCl4+ 2H2+ Si = 4SiHCl3 (2.13)

This hydrogenation reaction produces approximately 20% trichlorosilane at

500C and 35 bar with a 1:1 ratio of SiCl4 to H2 in one pass through

metallurgical grade silicon in a fluidized bed reactor.

In spite of its widespread and dominant position in the industry, the Siemens process as

described above suffers from the following disadvantages:

- High energy consumption, over 90% of the input power is lost to the cold walls

of the reactor.

- Two power supplies and preheating of the seed rods are normally required

because the high-resistivity (~230,000 ohm cm) seed rods require very high

power supplies and high initial power rates to heat the rods.

- Hot spot formation and filament burn out may occur.

- The process is operated batchwise.

- Large amount of by-products need to be handled or recycled.


34/231


35/231

21

suitable for single-crystal manufacturing by the floating zone (FZ) method. The

disadvantage of the monosilane-based process is the high cost since additional steps are

required to convert trichlorosilane to monosilane. Each recycling of unsuitable

chlorosilanes yields only a small percentage of the suitable silane.

2.2.3 Ethyl Corporation Process

The Ethyl Corporation process is, by comparison with the Siemens and the Union

Carbide processes, revolutionary in all aspects except for the concept of purifying and

decomposing a volatile silicon compound by pyrolysis. The first radical change was thechoice not to use metallurgical grade silicon as the primary raw material for silane. The

idea was to make use of silicon fluoride, which is a waste by-product of the huge

fertilizer industry. Tens of thousands of tones of silicon fluoride every year are

available. This is potentially a very low-cost starting material. Silicon fluoride is

hydrogenated to monosilane by metal hydrides such as lithium aluminium hydride or

sodium aluminium hydride.

2 H2+ M+ Al = AlMH4, M being Na or Li (2.17)

SiF4+ AlMH4= SiH4+ AlMF4 (2.18)

After distillation, monosilane SiH4is thermally decomposed to polysilicon as described

by (2.16) while AlMF4is believed to find application in the aluminium industry, making

it a valuable saleable product. However, to realize this process, Ethyl Corporation

introduced a second radical change, not using static silicon seed rods in a bell-jar reactorbut dynamic silicon seed spheres in a fluidised bed sustained by a gas stream of silane

and hydrogen. A schematic representation of a fluidised bed reactor is given in Figure

2.3. The fluidised bed reactor offers significant advantages compared to the bell-jar

reactor. Most of the shortcomings identified for the Siemens process are then


36/231

22

eliminated. The energy losses and hence the energy consumption are considerably

reduced because the decomposition takes place at a lower temperature. Another

advantage is that large reactors may be constructed and operated continuously, reducing

further the capital and operating costs [2-10]. The end products are small granules of

polysilicon that may present some advantages (e.g. when continuous feeding in customer

process is requested) or disadvantages (e.g. not usable for direct float zone

crystallisation). A disadvantage of the process is the generation of powder due to

homogeneous decomposition of SiH4 in the free reactor space and the hydrogen

absorption into the polysilicon deposition layer.

Figure 2.3. A schematic representation of a fluidized bed reactor for polysilicon

production [2-11].

All mentioned processes have several common properties. All present processes for

production of pure silicon are energy intensive, involve toxic chemicals and are

relatively low capacity that increases cost per kg for produced pure silicon.


37/231

23

2.3 Metallurgical Techniques

The disadvantages present in current processes for the production of pure silicon could

be overcome by implementation of a set of metallurgical techniques. The most

recommended metallurgical techniques for silicon refining are acid leaching, alloying,

directional solidification, reactive gas blowing, high temperature vacuum treatment or

slagging.

2.3.1 Acid Leaching

Most of the impurities in silicon lie in the inter-crystalline boundaries or areas with grain

defects [2-12, 2-13]. An efficient route of removing the majority of the impurities is to

crush the silicon and treat the resulting powder. The particle size of the silicon prepared

for acid leaching is from 50 to 70m. The dissolution rate of the silicon itself is very

low in most of the leaching reactants [2-14]. Various acids like HCl, HF, H2SO4 and

HNO3or their combinations are used to dissolve silicon. Acids dissolve impurities such

as silicides of iron, calcium and magnesium as well as other impurities. The extent of

impurities removal depends on the particle size of the silicon and the duration of acid

leaching. However there are no reports about efficient removal of boron, phosphorous

and copper by the acid leaching technique.

2.3.2 Alloying

Very little work is being done in alloying metallurgical silicon in order to increase its

purity. Some attempts have been made with aluminum as the alloying element [2-15, 2-

16]. Work conducted on impurity elimination by alloying with Al proved to be partially

successful, especially for boron removal [2-15, 2-17]. However, Al is one of the most


38/231

24

efficient electron donors and is detrimental for Si solar cells. Furthermore, Si retrieval

from the Al-Si alloy is expected to be difficult because of the relative high segregation

coefficient of Al in Si, and leaching of Al is difficult due to formation of the Al-oxide or

Al-hydrates that inhibit further dissolution. Also, the low density difference between Si

and Al makes gravity separation unfeasible. Iron is another potential Si purifier due to

the high-density difference between Si and Fe. However, Fe effective ionic radius is

higher than Cu and this reduces the Fe mobility in solid Si.

Recent discoveries show that Cu can be efficiently used as an alloying agent in the route

to Si purification. An interesting finding is the diffusion coefficient of Cu in Si [2-18].

The data shows that Cu is the fastest diffusing foreign atom in solid Si. Cu diffuses an

order of magnitude faster than B, Al or P, and this difference is even higher at lowertemperatures. Also, Cu properties like low activity coefficient in silicon for most

impurity elements, low effective ionic radius and low cost indicate high potential of

inheriting impurities from the silicon.

2.3.3 Directional Solidification

The solid solubility in silicon is extremely small for most elements; therefore directional

solidification is an efficient refining technique [2-19]. The effectiveness of directional

solidification can be estimated by the segregation coefficient value. The segregation

coefficient gives the relation between the concentration of the impurity atom in the

growing crystal and that of the melt. Values of the segregation coefficient are usually

much below one because impurity atoms usually stay in the melt. In other words, the

solubility of impurity atoms in the melt is larger than in the solid. Solid silicon shows a

very low solubility for most of the elements, much lower than liquid silicon, resulting ina strong segregation of these elements at freezing [2-20]. A typical example is iron that

shows a segregation coefficient of 8x10-6. Segregation coefficients for various

impurities in silicon are given in Table 2.1.


39/231

25

Table 2.1. Segregation coefficients for various impurities in silicon [2-21]

ImpuritySegregation

coefficientImpurity

Segregation

coefficient

Aluminium 2.0 x 10 -3 Molybdenum 4.5 x 10 -8

Antimony 2.3 x 10 -2 Nickel 1.0 x 10 -4

Arsenic 3.0 x 10-1 Niobium 4.4 x 10 -7

Bismuth 7.0 x 10 -4 Palladium 5.0 x 10 -5

Boron 8.0 x 10-1 Phosphorus 3.5 x 10-1

Carbon 5.0 x 10-2 Silver 1.7 x 10 -5

Chromium 1.1 x 10 -5 Tantalum 2.1 X 10 -8

Cobalt 2.0 x 10 -5 Tin 1.6 X 10 -2

Copper 4.0 x 10 -4 Titanium 2.0 X 10 -6

Gallium 8.0 x 10 -3 Tungsten 1.7 X 10 -8

Indium 4.0 x 10 -4 Vanadium 4.0 X 10 -6

Iron 8.0 x 10 -6 Zinc 1.0 X 10 -5

Magnesium 3.2 x 10 -6 Zirconium 1.6 X 10 -8

Manganese 1.3 x 10 -5

So far only the Czochralski, float-zone and heat exchange methods have been

experimentally tried to achieve efficient SGS from MGS. The success of refining by

directional solidification techniques depends on the initial chemical composition of the

MGS. Higher concentration of elements with high segregation coefficient reduces the

efficiency of the directional solidification process. Particularly, boron, phosphorous and

aluminum have high segregation coefficients; therefore, they are not easily removed

from the silicon. At the same time, they are considered as dopants in silicon for

electronic devices and an increased amount of these impurities significantly decrease the

photovoltaic properties of the silicon based solar cells. Currently, a combination of

purification processes by chemical methods and directional solidification has provided

the best results. Repeating the refining processes increases the purity but trades off in the

final costs for silicon.


40/231

26

2.3.4 Reactive Gas Blowing

Reactive Gas Blowing is one of the common techniques practiced by the silicon refining

industry to obtain a purity level of at least 99.99% [2-22] where can only be used to

refine elements that are more reactive than silicon, such as Al, Ca, Mg, Na... The

method could be used in combination with other refining techniques, either as the first

step of refining or before directional solidification. The gases used in silicon refining

industry are Cl2, O2, SiCl4, wet hydrogen and CO2or their combinations, usually diluted

with an inert gas. Cl2and wet hydrogen form low boiling point compounds with various

elements that are easily volatilized.

2.3.5 Slagging

A slagging technique is used to remove elements less noble than silicon. There have

been many attempts utilizing CaCO3, BaO, MgO, Al/SiO2, CaO/SiO2 or CaF2/SiO2 to

make a refining slag. The amount of the artificial slag is usually 1-5% of the amount of

silicon. Elements like Al, Ca and Mg are oxidized at a high rate. The degree of refining

is determined by the distribution equilibrium of the respective reaction (equations 2.19

to 2.22).

4 Al + 3 SiO2(SLAG)= 3 Si (LIQUID)+ 2 Al2O3(SLAG) (2.19)

2 Ca + SiO2 = Si (LIQUID)+ 2 CaO (SLAG) (2.20)

2 Mg + SiO2 = Si (LIQUID)+ 2 MgO (SLAG) (2.21)Si (LIQUID)+ O2= SiO2(SLAG) (2.22)

The impurity level reduction for some of the elements such as Ti, Mn, V and Al are

about one order of magnitude.


41/231

27

2.3.6 High Temperature Vacuum Refining

Vacuum refining is a technique used to purify and separate one liquid from another. It is

based on chemical volatility or the relative ease of which the molecules leave the surface

of a liquid. The successful application of vacuum refining techniques depends on

several factors such as the difference in boiling points of chemicals present, the size of

the sample and type of refining apparatus. Several reports claim success in removing

boron and phosphorous by vacuum refining at temperatures above the melting point of

silicon. However, the feasibility is limited due to requirements for temperatures above

1500o

C and vacuum at a level below 0.05atm [2-23, 2-24].

2.4 Refractory Material for Si Refining

One of the important items in refining metallurgical grade silicon is the refractory

material. It has to withstand the refining temperature at about 1550oC, minimize

contribution of impurities to the molten silicon, be compatible with silicon and type of

heating system, and be low cost. Since all these characteristics are important, a task of

the investigation was to evaluate crucible materials that can be used for refining

metallurgical grade silicon. Table 2.2 gives physical characteristics for several

commonly used refractory materials. The types of crucible materials taken into

consideration were mullite, alumina, silicon carbide, magnesia, graphite and glass.


42/231


43/231

29

Mullite has long been used as a refractory material. Its properties include:

- Good high temperature strength.

- Good thermal shock resistance.

- Excellent thermal stability.

- It has excellent stability in acid slags and is insoluble in most acids.

- Resistance to oxidation and furnace atmospheres.

- Resistance to abrasion.

- Good electrical resistivity.

The approximate limiting temperatures of use are 1725oC in air and 1600oC in vacuum.

The binary phase diagram of Al2O3and SiO2is given in Figure 2.4.

Figure 2.4. Al2O3-SiO2phase diagram where mullite is an intermediate compound

with ideal stoichiometry 3Al2O3-2SiO2. [2-25]


44/231


45/231

31

2.4.4 Silicon Carbide

Silicon carbide was originally produced by a high temperature electro-chemical reaction

of sand and carbon. Silicon carbide is composed of tetrahedra of carbon and silicon

atoms with strong bonds in the crystal lattice. This produces a very hard and strong

material. Up to 800C, silicon carbide is not attacked by any acids or alkalis or molten

salts. In air, silicon carbide forms a protective silicon oxide coating at 1200C. The

high thermal conductivity coupled with low thermal expansion and high strength gives

this material exceptional thermal shock resistant qualities. Silicon carbide can also be

used as an electrical conductor and has applications in resistance heating, flame ignitersand electronic components.

Table 2.5. Typical physical and mechanical properties of the carbide.

Property

Density (g/cm3) 3.21

Youngs Modulus (GPa) 207-483

Fracture Toughness (MPa.m-1/2) 3-5

Thermal Expansion Coefficient (x10-6/C) 4.3-5.6

Thermal Conductivity (W/m.K) 33-155

Silicon carbide chemical purity, resistance to chemical attack at temperature, and

strength retention at high temperatures has made this material very popular as wafer tray

supports and paddles in semiconductor furnaces where with little or no grain boundary

impurities maintains strength to very high temperatures, approaching 1600C with no

strength loss. The most beneficial properties of silicon carbide are:

High hardness (second only to diamond).

Low density.

Low porosity.


46/231

32

Good wear resistance in sliding and abrasive environments.

Excellent corrosion resistance in most chemical environments.

Low thermal expansion and high thermal conductivity leading to excellent

thermal shock resistance.

However, the high price and possible carbon contamination puts this material below

mullite as a possible refractory material for metallurgical silicon refining.


47/231

33

3 Impurity Detection in High Purity Silicon

There are a number of techniques for chemical composition measurements of pure

silicon such as Inductively Coupled Plasma - Mass Spectrometry (ICP-MS) and

Inductively Coupled Plasma Optical Emission Spectroscopy (ICP-OES) [3-1:4],

Atomic Absorption Spectrophotometry (AAS), Standard Titration technique (STT) [3-

5:8], Glow Discharge Mass Spectrometry (GDMS) [3-9:13], LECO-Combustion

Techniques [3-15:17], etc.

An effective technique for impurity detection in pure silicon is Inductively Coupled

Plasma (ICP). However, ICP is limited in that it is not capable of measuring carbon oroxygen. In addition, the measurement resolution is not sensitive enough to detect the

lower limit of detrimental elements like titanium, calcium, boron and phosphorus. In

order to provide a full analysis of silicon, the use of other measurement techniques such

as Glow Discharge Mass Spectroscopy (GDMS) and nitrogen/oxygen determination

(carried out on LECO Corporation equipment) is necessary. Table 3.1 shows

recommended techniques in order to determine exact chemical composition in pure

silicon.

Table 3.1. Most suitable measurement techniques for trace amounts of impurities

in pure silicon [3-18].

Element Impurity MeasurementCalcium GDMSCopper ICP-OES

Phosphorous GDMSBoron GDMS

Iron ICP-OES

Carbon LECO C-200Oxygen LECO TC-436

Aluminum ICP-OESNickel ICP-OES

Titanium GDMS


48/231

34

3.1 Inductively Coupled Plasma Mass Spectrometry (ICP)

ICP is a comprehensive technique able to analyze most elements from lithium through

uranium. It is an extremely sensitive technique with detection limits in the parts pertrillion (ppt) ranges for many elements in aqueous solutions. An approximate detection

limit for the ICP technique used at the University of Toronto is given in Figure 3.1. Its

high level of relative accuracy (1 to 2%) coupled with its sensitivity allows the analyst to

cover more than nine orders of magnitude in concentration. This has eliminated the

need to use several techniques to obtain a complete analysis. Recent technological

advances in Cool Plasma ICP have helped eliminate or reduce interferences caused by

the argon plasma thus allowing lower detection levels for Li, Na, K, Ca and Fe.

ApproximateDetection Limit:

Figure 3.1. An approximate detection limit for the ICP technique used at the

University of Toronto [3-19].


49/231

35

3.2 Atomic Absorption Spectrophotometry (AAS)

This is a technique for elemental analysis in liquids. Metallic species can be determined

in both organic and inorganic samples. It is a sensitive technique that can measure theconcentration of most elements at part - per - million (ppm) levels. If lower detection

limits are required, then a graphite furnace is used as the excitation source (GFAAS),

replacing the standard flame. GFAAS is utilized for ultra-trace analysis. It has detection

limits that exceed the conventional flame by several orders of magnitude. Figure 3.2

shows elements that can be detected by the atomic absorption spectrophotometry

technique.

Through the use of calibration curves, prepared from suitable standards, a high level of

accuracy and precision (1 to 3 %) is achieved for the flame AAS technique. Most

spectral interferences and matrix effects are overcome by utilizing well characterized

conditions and matrix matching of samples and standards. AAS has an average

accuracy which is suitable for reporting at the ultra-trace impurity level.

Figure 3.2. Elements determined by the AAS [3-20]


50/231

36

3.3 Standard Titration Technique (STT)

Titration is a common laboratory method of quantitative chemical analysis that can be

used to determine the concentration of a known reactant. Because volume measurementsplay a key role in titration, it is also known as volumetric analysis. A reagent, called the

titrant, of known concentration (a standard solution) and volume is used to react with a

solution of the analyte, whose concentration is not known in advance. Using a calibrated

burette to add the titrant, it is possible to determine the exact amount that has been

consumed when the endpoint is reached. The endpoint is the point at which the titration

is complete, as determined by an indicator. Standard titration technique is the most

suitable for analysis of the copper in Cu-Si alloys.

3.4 Glow Discharge Mass Spectrometry (GDMS)

In Glow-Discharge Mass Spectrometry (GDMS), the sample to be analyzed forms the

cathode in a low pressure (~ 100 Pa) gas discharge or plasma. Argon is typically used asthe discharge gas. Argon positive ions are accelerated towards the cathode (sample)

surface with energies from hundreds to thousands of eV resulting in erosion and

atomization of the upper atomic layers of the sample. Only the sputtered neutral species

are capable of escaping the cathode surface and diffusing into the plasma where they are

subsequently ionized. The atomization and ionization processes are thus separated in

space and time, which appears to be a keystone for simplified calibration, quantification

and the near matrix independence of this technique. Strengths of GDMS Analysis is full

periodic table coverage (except H), sub-ppb to ppt detection, minimal matrix effects and

linear and simple calibration. However, GDMS technique is very expensive.

Documents

Solubility of Copper in Silicons