SILICON REFINING THROUGH CHEMICAL VAPOUR DEPOSITION
by
Mark (Xiang) Li
A thesis submitted in conformity with the requirements for the degree of Master of Applied Science Graduate Department of Materials Science and Engineering University of Toronto
© Copyright by Mark Xiang Li 2008
SILICON REFINING THROUGH CHEMICAL VAPOUR DEPOSITION
Master of Applied Science 2008 Mark Xiang Li Department of Material Science and Engineering University of Toronto
ABSTRACT
Currently the cost of solar grade silicon accounts for approximately one third of the total solar cell cost, therefore a new silicon refining process is being proposed with the goal of lowering the cost of producing solar grade silicon.
In this new process, Si-Cu alloys were used as the silicon source. One to one molar ratio
H2-HCl gas mixtures were used as transport agents to extract Si out from the Si-Cu alloy at about 300-700oC, with following reaction taking place:
Si+3HCl(g)=HSiCl3(g)+H2(g)
While at about 1000-1300oC, pure Si deposits onto a hot silicon rod according to:
Si+3HCl(g)=HSiCl3(g)+H2(g)
The role of the copper in the alloy was to trap impurities in the Si and catalyze the gas solid reaction. A study on determining the rate limiting step and impurity behavior was done. A possible silicon extraction reaction mechanism was also addressed.
ii
ACKNOWLEGEMENTS
I would like to express my deep and sincere gratitude to all those who gave me the possibility to complete this thesis. I am deeply indebted to my supervisor Prof. Torstein Utigard whose wide knowledge and his logical way of thinking have been of great value for me. His understanding, encouraging and personal guidance have provided a good basis for the present thesis.
My warm thanks are due to Anton, Horazio, Gabriel and Alex for valuable advices and friendly help.
I would also like to thank ARISE Technologies, as well as University of Toronto Open Fellowship for the financial support.
I would like to give my special thanks to my family whose patient love enabled me to complete this work.
iii TABLE OF CONTENTS
1. INTRODUCTION ...... 1 2. OBJECTIVES...... 4 3. LITERATURE REVIEWS ...... 5 3.1 ELECTRONIC GRADE SILICON PRODUCTION ...... 5 3.1.1 Metallurgical Grade Silicon Production ...... 6 3.1.2 Trichlorosilane Production...... 7 3.1.3 Siemens Reactor...... 7 3.2 PROCESSES FOR PURIFICATION OF SOLAR GRADE SILICON ...... 9 3.3 Cu-Si ALLOYS AS SILICON SUPPLY SOURCES...... 9 3.4 DEPOSITION VARIABLES...... 14 3.4.1 Cu-Si Alloy Compositions...... 15 3.4.2 Deposition Surface Areas ...... 17 3.4.3 Surface Areas of Cu-Si Alloys...... 19 4. EXPERIMENTAL CONDITIONS AND PROCEDURE...... 23 4.1 EXPERIMENTAL CONDITION DETERMINATION...... 23 4.2 REACTOR DESIGN ...... 27 4.2.1 Reactor ...... 27 4.2.2 Cu-Si Alloys...... 29 4.2.3 Silicon Starting Rod...... 30 4.2.4 Power Supply...... 32 4.2.5 Temperature Measurement Device...... 33 4.3 EXPERIMENTAL PROCEDURES...... 34 4.3.1 Prior to Experiments ...... 34 4.3.2 During the Experiment...... 35 4.3.3 After the Experiment...... 36 4.4 EXPERIMENTAL UNCERTAINTIES ...... 36 5. SILICON DEPOSITION CALCULATION...... 37 6. RATE LIMITING STEP...... 42 6.1 BY PRODUCTS FORMATION ...... 44 6.2 MASS FLOW RATE...... 44 7. PRIMARY SILICON REACTION PHASE...... 46 7.1 EXPERIMENT WITH 30, 50 AND 75 Cu-Cu ALLOYS...... 48 7.1.1 Experimental Details...... 49 7.1.2 Results and Discussions...... 50 7.1.2.1 Silicon Recovery and Cu-Si Alloy Weight Losses...... 50 7.1.2.2 Cu-Si Alloys After The Experiment ...... 53 7.1.3 Conclusions...... 54 7.2 REPEATED EXPERIMENT WITH 30, 50 AND 75 WT%Si-Cu ALLOYS...... 55 7.2.1 Experimental Details...... 56 7.2.2 Results and Discussions...... 57 7.2.2.1 Silicon Recovery and Cu-Si Alloy Weight Losses...... 57 7.2.2.2 Si Deposits ...... 58 7.2.2.3 Cu-Si Alloys After The Experiment ...... 59
iv 7.2.3 Conclusions...... 60 7.3 EXPERIMENT WITH 30 AND 50WT%Si-Cu ALLOYS ...... 61 7.3.1 Experimental Details...... 61 7.3.2 Results and Discussions...... 63 7.3.2.1 Silicon Recovery and Cu-Si Alloy Weight Losses...... 63 7.3.2.2 Silicon Deposit...... 64 7.3.2.3 Cu-Si Alloy After The Experiment...... 64 7.3.3 Conclusions...... 66 7.4 SILICON DIFFUSION...... 67 7.4.1 Density of Alloys Before and After the Experiment ...... 67 7.4.1.1 Experimental Setup...... 67 7.4.1.2 Results...... 68 7.4.2 Compositions of the Alloy Core ...... 69 7.4.2.1 Experimental Setup...... 69 7.4.2.2 Results...... 69 7.5 CONCLUSIONS AND FUTURE WORK ...... 70 8. ALLOY CRUMBLES AND SILICON EXTRACTION MECHANISMS...... 71 8.1 ALLOY CRUMBLES TESTS...... 71 8.1.1 Experimental Details...... 71 8.1.2 Observations and Results...... 72 8.1.3 Crumbles Microstructures...... 75 8.2 COMPOSITION OF THE CRUMBLES...... 77 8.2.1 Experimental Setup...... 77 8.2.2 Results...... 78 8.3 PROPOSED SILICON EXTRACTIOIN MECHANISM...... 79 8.4 CONCLUSIONS...... 80 9. IMPURITY STUDY...... 81 9.1 SOURCE OF IMPURITIES ...... 82 9.2 IMPURITIES IN THE Cu-Si ALLOYS...... 84 9.2.1 Experimental Setup...... 85 9.2.2 Results...... 85 10. CONCLUSIONS AND FUTURE WORK ...... 88 11. REFERRENCES...... 90 Appendix A: Chemical Analysis On Cu, MG-Si, Cu-Si Alloys, EG-Si and Deposited Si ...... 92 A.1 Metallurgical Grade Silicon...... 92 A.2 Copper Used for Alloying with MG-Silicon...... 93 A.3 Electronic Grade Silicon ...... 94 A.4 Impurities in the 30wt%Si-Cu alloy (Area 1) ...... 95 A.5 Impurities in the 30wt% Si-Cu alloy (Area 2) ...... 96 A.6 Impurities in the 30wt% Si-Cu alloy (Area 3) ...... 97 A.7 Impurities in the 30wt% Si-Cu alloy (Area 4) ...... 98 A.8 Impurities in the 30wt% Si-Cu alloy (Area 5) ...... 99 A.9 Impurities in the 30wt% Si-Cu alloy (Area 6) ...... 100 Appendix B. The GASEOUS SPECIES OF ELEMENTS Al, B, Si, Fe, P, Ti and Mn. 101 Appendix C. silicon refining efficiency...... 102
v LIST OF FIGURES
Figure 1.1 Energy Flux comparison ...... 1 Figure 1.2 U.S. Electricity Generation by Energy Source, 2004 ...... 2 Figure 1.3 Cost comparison between different grades of silicon, 2003 ...... 2 Figure 1.4 Predicted solar grade silicon demand and supply...... 3 Figure 3.1 Schematic drawing of electronic grade silicon production ...... 5 Figure 3.2 A typical plant for the production of silicon metal...... 6 Figure 3.3 Siemens Reactor ...... 8 Figure 3.4 SEM illustration of 25wt%Si-Cu Alloy. Dark phase is the primary Si phase, lighter phase is the Cu-Si matrix...... 10 Figure 3.5 schematic drawing of experimental reactor...... 11 Figure 3.6 Cu-Si Binary Phase Diagram...... 12 Figure 3.7 Relative contribution of the different mechanisms to the overall purifications ...... 14 Figure 3.8 Experimental set-up in Ebner ...... 21 Figure 3.9 Different alloys stacking up...... 21 Figure 4.1 Partial pressure of possible gaseous species from H, Cl, and Si...... 24 Figure 4.2 Sums of the equilibrium partial pressures of silicon gaseous species (HCl-H2- Si mol ratio is 1:1:50) ...... 25 Figure 4.3 Sums of the equilibrium partial pressures of Al, Fe, B, Ti, Mn and P gaseous species...... 27 Figure 4.4 schematic of experimental set-up ...... 27 Figure 4.5 DC Power Supply...... 28 Figure 4.6 Cast Cu-Si alloys in puck and square shapes ...... 29 Figure 4.7 Alloys in rectangle shape...... 30 Figure 4.8 Electronic grade silicon starting rod...... 30 Figure 4.9 Actual reactor set-up...... 31 Figure 4.10 Power supplies (a) Main DC power supply (b) External heating AC power supply...... 32 Figure 4.11 M770S two color infrared sensor ...... 34 Figure 5.1 Basic Experimental Schematic for Si Deposition Calculation...... 37 Figure 6.1 Silicon diameter vs time ...... 43 Figure 6.2 The weight and diameter of the rod in the first few hours reaction ...... 43 Figure 6.3 Diameter vs time with various pumps...... 45 Figure 7.1 Cu-Si phase diagram...... 46 Figure 7.2 SEM pictures of Cu-Si alloys with various Si compositions ...... 47 Figure 7.3 The Cu-Si alloys with 30, 50 and 75 wt%Si used in the experiment...... 48 Figure 7.4 Eutectic liquid drops on the 50wt%Si alloy ...... 52 Figure 7.5 Silicon deposits after 5.5h reaction ...... 52 Figure 7.6 Cu-Si alloys after 5.5 h reaction. Alloys with higher silicon content are less crumbly and swelling...... 53 Figure 7.7 The 50wt%Si alloy after 5.5h reaction...... 54 Figure 7.8 The Cu-Si alloys with 30, 50 and 75 wt%Si used in the experiment...... 55 Figure 7.9 Deposited silicon (a) front view (b) top view (c) bottom view ...... 58
vi Figure 7.10 Sketch of boundary layer created around silicon rod due to a fluid flow...... 59 Figure 7.11 The Cu-Si alloys after 5.5 h reaction...... 60 Figure 7.12 The Cu-Si alloys with 30 and 50wt%Si used in the experiment...... 61 Figure 7.13 Silicon deposit ...... 64 Figure 7.14 The Cu-Si alloys after the experiment...... 65 Figure 7.15 The disintegrated 30wt%Si alloy...... 65 Figure 7.16 The 30wt%Si-Cu alloy core ...... 66 Figure 7.17 The experimental setup for measuring volume of the alloy core ...... 67 Figure 7.18 Schematic of cutting the alloy core ...... 69 Figure 8.1 Macroshots of the alloys during and after the experiments...... 73 Figure 8.2 The 30 and 75wt%Si Alloy cores after the 24 hr experiment ...... 75 Figure 8.3 Sintered crumble layer from the 30wt%Si alloy ...... 76 Figure 8.4 SEM of the crumble layer from the 30wt%Si alloy ...... 76 Figure 8.5 EDX analysis result of a crumble...... 77 Figure 8.6 Diffractogram of the crumbles from the 30wt%Si-Cu alloy...... 78 Figure 8.7 Illustration of the proposed silicon extraction mechanisms ...... 79 Figure 9.1 Cu-Si matrix phase which may retain most impurities ...... 84 Figure 9.2 The 30wt%Si alloy and its slag ...... 85 Figure 9.3 EDX analysis on the boundary between the Si dendrites and the matrix...... 86 Figure 9.4 EDX analysis on the needle structure which is silicon eutectic ...... 86
vii LIST OF TABLES
Table 3.1 Test conditions and measurements before the experiments for Cu-Si alloys with various compositions ...... 15 Table 3.2 Measurements after the 24 h experiments ...... 16 Table 3.3 Test conditions and measurements before experiment for different deposition surface areas...... 17 Table 3.4 Measurements after the experiments for different deposition surface areas .... 18 Table 3.5 Test conditions and measurements before the experiments for the different alloy surface areas ...... 19 Table 3.6 Measurements after experiment for different Cu-Si surface areas ...... 20 Table 4.1 purity level in MG-Si...... 26 Table 4.2 Reactor Tube Dimension ...... 28 Table 4.3 Purity level of silicon staring rods ...... 31 Table 6.1 Measurements after the 10 hrs experiments using different number of pumps 45 Table 7.1 Test conditions and measurements before the experiment...... 49 Table 7.2 Measurements after 5.5 h of experiment for Cu-Si alloy ...... 50 Table 7.3 Test conditions and measurements before the experiment...... 56 Table 7.4 Measurements after 5.5 h of experiment for Cu-Si alloy ...... 57 Table 7.5 Test conditions and measurements before the experiment...... 62 Table 7.6 Measurements after 5.5 h of experiment for Cu-Si alloy ...... 63 Table 7.7 the percentage of Si weight losses from silicon in the alloy for different experiments...... 64 Table 7.8 Densities of the 30wt%Si alloy and its core before and after the reaction...... 68 Table 7.9 Si content in the alloy core ...... 70 Table 8.1 Tests conditions and measurements before the experiments for the different Cu- Si alloys...... 72 Table 8.2 Test conditions and measurements after the experiments for the different Cu-Si alloys...... 74 Table 8.3 Measurements of the crumbles before and after the 24 hr experiment...... 78 Table 9.1 Impurity analysis results by GDMS on the silicon deposit using the 30wt%Si alloy...... 81 Table 9.2 Impurity analysis results by GDMS on the silicon deposit using the 75wt%Si alloy...... 81 Table 9.3 Average ICP analysis results on impurities in the 30wt%Si alloy ...... 82 Table 9.4 Impurity analysis result by GDMS on MG-Si...... 83 Table 9.5 Impurity analysis result by GDMS on Cu used for alloying with MG-Si ...... 83
viii 1. INTRODUCTION
People are more than ever interested in solar energy. Reasons for this are:
• The sun provides a virtually unlimited supply of solar energy • No green house gas emissions • Silicon, the backbone material used in solar cells, is one of the most abundant elements on Earth. • The solar energy received by the earth far exceeds all other potential energy sources (Fig. 1.1)
Flows of Various Energies per Year
10000
1000
100
Energy(ZJ) 10
1
0.1 Solar Energy Wind Biomass World Energy Consumption
Figure 1.1 Energy Flux comparison [1-5]
In spite of these benefits, solar cells only produced around 0.018% of the total electric energy in the United States in 2004 (Fig. 1.2).
1
Figure 1.2 U.S. Electricity Generation by Energy Source, 2004 [6]
The reason that solar energy contributes only a small portion of electric generation is that the cost of producing electricity by utilizing solar cells is very high. The high cost of solar cell modulus is partly due to high purity requirement and high energy demand for refining solar grade silicon. According to silicon impurity levels, silicon can be classified into three types: metallurgical grade silicon (MG-Si), solar grade silicon (SG-Si) and electronic grade silicon (EG-Si). Solar grade silicon is the one used for solar cells.
10000
1000
Metallurgical 100 Grade Silicon ppm) 10 y Level (
it 1 Impur 0.1 Solar Grade Silicon
0.01
Electronic 0.001 Grade Silicon 0 20406080100 Cost (US$/kg)
Figure 1.3 Cost comparison between different grades of silicon, 2003 [7]
2 The cost of solar grade silicon is about one third of the total solar cell module cost. Even though the price of solar cell modules is as expensive as $4/watt, in recent years, with more concerns about global warming and energy crisis, more companies have entered into the photovoltaic industry [8].
Solar grade silicon presently mainly comes from off-specification electronic grade silicon. With the explosive growth of the photovoltaic industry, market demand for SG-Si has far exceeded the off-spec EG-Si supply. As a result, the market price of SG-Si has already skyrocketed from 30US$/kg to 260US$[9].
Figure 1.4 Predicted solar grade silicon demand and supply
The SG-Si shortage problem cannot economically be solved by expanding current EG-Si plants. There are so many purification processes involved in EG-Si plants, and purity level of the produced silicon far exceeds the needs of solar cells. Thus, a newly proposed chemical vapor deposition (CVD) process utilizing a Cu-Si alloy as Si feeds is being developed to produce silicon with sufficient purity level in a more cost-effective manner. The role of Cu in the alloy is to trap impurities and catalyze the gas-silicon reaction. A feasibility study of the process was successfully demonstrated at the lab prototype stage [7][10]. However, the mechanism of the reaction and behaviour of the impurities from MG-Si are still unclear.
3 2. OBJECTIVES
The proposed CVD process is a modification of the standard Siemens process, which is used to produce electronic grade silicon. After trichlorosilane (HSiCl3, or TCS) is produced and refined, it is passed into a Siemens reactor and then decomposes into high purity silicon that will deposit on electrically heated silicon rods. The new proposed silicon refining process combines the production of trichlorosilane (HSiCl3, or TCS) and silicon deposition in one single reactor. The motivation of this combination is to cut the cost of the silicon refining process, while still maintains sufficient purity levels.
This thesis focuses on the investigation of rate limiting steps, extraction mechanisms and impurity behaviour of this simple reactor silicon refining process.
4 3. LITERATURE REVIEWS
3.1 ELECTRONIC GRADE SILICON PRODUCTION
A schematic of the electronic grade silicon production is illustrated in Figure 3.1. [11]
Figure 3.1 Schematic drawing of electronic grade silicon production
5 3.1.1 Metallurgical Grade Silicon Production
The earth’s crust contains about 28% silicon bound as SiO2 silicates and a large fraction of this exists as relatively pure SiO2. To produce high purity metallurgical grade silicon,
carbon is utilized to reduce SiO2 in arc furnaces with idealized carbothermal reaction shown in Equation 3.1:
SiO2 + 2C = Si + 2CO [3.1]
A silicon metal plant that is designed based on above equation is shown in Figure 3.2. [12]
Figure 3.2 A typical plant for the production of silicon metal
6 o The hottest area in the furnace is about 2000 C. At this high temperature SiO2 is reduced to molten silicon by reacting with carbon, which is tapped from the furnace through a taphole at the bottom and refined by slag treatment or gas purging. [12]
3.1.2 Trichlorosilane Production
The metallurgical grade silicon produced from arc furnaces reacts with HCl gases in fluidized bed reactors at 300oC to form a silicon-containing gas [10]:
300°C Si(s) + 3HCl( g) ⎯⎯→⎯ HSiCl3(g ) + H 2( g) [3.2]
Approximately 90% of the gas produced from the fluidized bed reactor is HSiCl3, the rest
being mainly SiCl4, which can be converted to HSiCl3 by reacting with H2 in the presence of MG-Si and copper catalyst:
Cu 3SiCl4( g ) + 2H 2( g ) + Si(s) ⎯⎯→ 4HSiCl3(g ) [3.3]
The produced HSiCl3 can be purified to a very high level by distillation, since chlorides formed by other metal impurities have different boiling points. After the purification process, pure HSiCl3 gas is ready for chemical vapor deposition (CVD) process.
3.1.3 Siemens Reactor
The Siemens process is a kind of chemical vapor deposition (CVD) process. HSiCl3 is flushed into a quartz bell jar reactor shown in Figure 3.3 and reduced by H2 to form bulk polysilicon:
1100°C HSiCl3( g) + H 2( g) ⎯⎯→⎯ Si(s) + 3HCl( g) [3.4]
7 The reduced polysilicon is deposited on a U-shaped silicon filament which is resistively heated to 1100oC.
The major by-product of this process is SiCl4, which is produced by:
SiHCl3(g ) + HCl(g ) → SiCl4(g ) + H 2(g ) [3.5]
SiCl4 can either be recycled back to the fluidized bed reactor forming HSiCl3 again by Equation 3.3 or be used as a silicon source for deposition thin-films of silicon.
Figure 3.3 Siemens Reactor [13]
The polysilicon produced from the Siemens reactor requires further processing, which is called Czochralski (Cz) or Floating Zone (Fz) method for electronic device applications.
8 Cz and Fz are the standard methods employed in the semiconductor industry for single- crystalline silicon growth from polysilicon. [13]. During the whole refining process, the off-specification and chopped off silicon will be used in the photovoltaic industry.
3.2 PROCESSES FOR PURIFICATION OF SOLAR GRADE SILICON
Silicon purification processes can be classified into three basic types:
1. Adaptations and modifications of the Siemens process. Formed of volatile silicon compounds, such as silane or chlorosilanes, which can be purified by distillation and from which high-purity silicon is recovered by pyrolysis.
2. Unconventional chemical pathways. Impurities are removed from soluble silicates or organosilanes, which are then reduced to silicon.
3. Metallurgical routes. Combination of directional solidification, slagging, reactive gas blowing, vacuum degassing, or acid Leaching. [14]
The purity specification required for silicon solar cell can be achieved without a doubt by a Siemens type process, however it is energy intensive and costly.
3.3 Cu-Si ALLOYS AS SILICON SUPPLY SOURCES
Research shows that Cu-Si alloys composed of a primary silicon phase, which is embedded in a Cu-Si matrix (see Figure 3.4). Cu-Si matrix phase can function as a “filter” to trap the impurities from metallurgical grade silicon. With the trapped impurities inside the alloy, pure HSiCl3 is able to form inside the reaction chamber. Due to the high purity level of the TCS formed, the distillation process is not required. Therefore, it should be possible to combine TCS formation and Si deposition in one single reactor. [15]
9
Figure 3.4 SEM illustration of 25wt%Si-Cu Alloy. Dark phase is the primary Si phase, lighter phase is the Cu-Si matrix.
In the experiments conducted by Tejedor et al [15], hypereutectic Cu-Si alloys containing 25wt% Si were used as the silicon source. The reaction condition was similar to the Siemens process. The reactor was first evacuated to a pressure of 30 to 50 Torres, and
then filled with gaseous HCl and H2. A graphite plate filament (65mm × 25mm ×1mm) was resistively heated to 1050-1300oC and the Cu-Si alloy slabs were radiatively heated to 600-700o. The experimental set-up is showed in Figure 3.5
10
Figure 3.5 schematic drawing of experimental reactor. [15]
In her work, the silicon deposition rate was around 0.61mm/hr. She claimed that the silicon deposition area was the rate limiting factor. From the Cu-Si phase diagram illustrated in Figure 3.6, it is seen that the η phase Cu-Si matrix is Cu3Si. [16][17] The purposes of alloying MG-Si with Cu are to catalyze the formation of TCS from silicon phase and to trap impurities in the Cu rich eutectic phase.
11
Figure 3.6 Cu-Si Binary Phase Diagram [16]
Many impurities exist in MG-Si, such as Al, B, Ba, Ca, Cr, Fe, Mg, Mn, Mo, Ni, P, Ti, V and Zr. Tejedor and Olson explained different mechanisms to the overall purification. [15] These impurities can be classified into four groups: impurities retained in slag formation during casting Cu-Si alloys; those retained very efficiently by the Cu3Si phase; those
12 retained with medium efficiency; and those incorporated almost fully into the vapor phase.
Impurities effectively retained by slag formation of Cu2O/SiO2 were Al and Ca.
Impurities retained by the Cu rich alloy with high efficiency were B, Ba, Ca, Cr, Fe, Mg, Mo, and Ni. The lower concentration of these elements found in the transported silicon was claimed to be due to their lower diffusion coefficients in the η-Cu3Si phase.
Less retained impurities are P, Ti, V and Zr. All of these impurities can form chlorides. However, the formation of chlorides on the alloys is favored at high alloy temperatures. Hence the depositions of those impurities onto the Si filament where the temperature is even higher than at the alloys were avoided.
Poorly retained impurities, Al and Mn, reacted with HCl to form AlCl3 and MnCl2, which condensed on the cold wall of the reactor due to their high boiling points [15]
The mass balance chart shown in Figure 3.7 supported Tejedor and Olson’s explanation. In conclusion, the work done by Tejedor et al demonstrated the possibility of using Cu-Si alloys as a Si source for refining MG-Si to SG-Si.
13
Figure 3.7 Relative contributions of the different mechanisms to the overall purifications [15]
3.4 DEPOSITION VARIABLES
Fan [10] conducted similar experiments which adapted Tejedor’s set-up to investigate variables affecting the silicon deposition. Studying those variables will help find out the rate limiting step in the new CVD process. Deposition rates in a traditional Siemens reactor can be increased by: increasing rod temperature; increasing TCS mole ratio; increasing the flow rate of TCS over the Si rod; increasing deposition area; increasing HCl removing rate over Si rod comparing to TCS and increasing pressure of the reactor. [18] However, TCS in the Siemens reactor is produced in a prior step, so the production of TCS is not a concern in the reactor. But in Fan’s reactor, which combined both formation of TCS and deposition of Si, the production of TCS from Cu-Si alloys has to be considered too. Fan suggested that the Cu-Si alloy composition and its surface area also influence the deposition rate. However, it has to be bear in mind that the combined implementation of these variables is complex. Since by changing one variable, the other
14 variables usually will be affected. And some changes may alter the chemistry of chlorides, such as formation of dichlorosilane, which will affect the deposition rate.
3.4.1 Cu-Si Alloy Compositions
Four tests with similar reaction conditions except composition of the Cu-Si alloys were conducted. Details about specifications of tests were listed in Table 3.1
Table 3.1 Test conditions and measurements before the experiments for Cu-Si alloys with various compositions [10] Test C1 Test C2 Test C3 Test C4 Si content (wt%) 15 25 30 50 Length (cm) 15.5 15.1 15.7 15.2 Si Average 5.0 4.86 5.00 4.95 Starting Diameter (mm) Rods Mass (g) 6.95 6.70 7.11 6.62 Si dendrite Si dendrite Si dendrite Solid Microstructure in Cu3Si in Cu3Si in Cu3Si solution Matrix Matrix Matrix Cu-Si Average Alloys 4.38 4.16 4.34 4.36 Thickness (mm) Length (mm) 83 95 105 104 Mass (g) 145.80 112.39 124.08 84.58 Experimental Duration (h) 24 24 24 24
The Cu-Si alloy used in Test C1 was a hypoeutectic alloy whose microstructure is different from that of hypereutectic alloys used in Tests 2-4.
The amount of silicon recovered from the Cu-Si alloy is shown in Table 3.2. The results are normalized to per length of silicon deposited so that they are comparable.
15 Table 3.2 Measurements after the 24 h experiments [10] Test C1 Test C2 Test C3 Test C4 Deposition 10.4 10.6 10.5 10.5 Length (cm) Average 8.47 8.20 7.45 11.23 Diameter (mm) Diameter 3.48 3.34 2.45 6.28 Gained (mm) Diameter Si Growth Rate 0.145 0.139 0.102 0.262 Starting (mm/h) Rods Mass (g) 13.54 14.21 13.39 23.52 Gained (g) 6.59 7.51 5.29 16.89 Deposition Rate 0.27 0.31 0.22 0.70 (g/h) Deposition Rate/Length 0.026 0.030 0.021 0.067 (g/h/cm) Cu-Si Mass (g) 138.50 102.02 115.67 65.76 Alloys Loss (g) 7.3 10.37 8.41 18.82
It is seen that the Si deposition rate utilizing hypereutectic alloys may depend on the composition of alloys. It appears that the higher Si wt%, the higher the deposition rate. Since with higher Si wt%, alloys contain more Si primary phases, which in turn may suggest the TCS was formed from the Si primary phase. However, because of the result of Test C3, which shows a slower rate than that of Test C2, the influence of Cu-Si alloys compositions on the deposition rate still needs to be verified.
16 3.4.2 Deposition Surface Areas
Three tests conducted by Fan were used to demonstrate the effect of deposition surface area on the silicon deposition rate. Silicon rods with different lengths but same diameter were used in these experiments. The details about the silicon rods and Cu-Si alloys for each test are listed in Table 3.3.
Table 3.3 Test conditions and measurements before experiment for different deposition surface areas Test L1 Test L2 Test L3 Intermediate Testing Condition Long rod Short rod rod Length (cm) 13.2 10.5 8.5 Si Starting Average 4.98 4.99 4.97 Rods Diameter (mm) Mass (g) 6.06 4.81 3.72 Average 9.6 9.7 N/A Cu-Si Thickness (mm) Alloys Length (mm) 94 87 106 Mass (g) 402 388 439 Experimental Duration (h) 27.5 51.5 69
The alloys used were 30wt%Si-Cu alloy. The temperature of the reactor was kept between 560 and 660oC.
The measurements of the silicon rods and Cu-Si alloys after the experiments are listed in Table 3.4. Fan suggested that the deposition rate on a weight basis is linearly to the silicon rod length and the deposition rate per unit deposit length did not vary significantly, which may suggest that TCS production is not a rate limiting step. Since the amount of TCS arrived on the silicon rod surface per unit length was the same.
17 Table 3.4 Measurements after the experiments for different deposition surface areas Test L1 Test L2 Test L3 Deposition Length (cm) 9.0 6.5 4.3 Diameter (mm) 12.38 16.17 17.50 Gained (mm) 7.4 11.2 12.5 Diameter growth rate 0.269 0.217 0.181 (mm/h) Si Rod Mass (g) 25.49 30.80 24.60 Gained (g) 19.4 26.0 20.9 Deposition Rate (g/h) 0.69 0.50 0.30 Deposition Rate/ Length 0.077 0.077 0.070 (g/h/cm) Mass (g) 380.2 407.5 364.2 Cu-Si Alloy Loss (g) 22 32 24 Experimental Duration (h) 27.5 51.5 69
However, the statement made by Fan is not completely accurate. The deposition surface area kept increasing with reaction time, since more and more Si deposited on the Si starting rod. If the TCS production is not rate limiting step and the supply of TCS is always enough for deposition, the diameter growth rate (or layer growth rate) should be the same. By utilizing Si rods as deposition filaments, with increasing diameter of the rod, the deposition rate on a weight basis should increase. Fan’s after experiment measurements clearly showed that the diameter growth rates in the three tests were not the same. The diameter growth rate in Test L1 was the highest among the three tests. Assuming the amount of TCS was the same at the beginning of the three tests (this is a reasonable assumption, since the amount of initial HCl/H2 mixture was the same in the three test as well as the amount of Cu-Si alloys), with increased reaction time, the production of the TCS would decrease because the silicon in the charged alloys kept being consumed in this batch process. When the TCS consumption rate from hot silicon rod side was faster than the TCS formation rate from the Cu-Si alloys, due to keep increasing deposition surface area, the deposition rate started to decrease. If the reaction still continued running after at, the deposition rate would further decrease. Even though the length of the Si rod used in Test L1 was longest, which in turn means largest surface area, the TCS consumption in Test L1 was faster than for the other two tests, but its reaction time was the shortest, which might lead to its largest layer deposition rate.
18 Therefore, TCS production might be the rate limiting step after sometime when the TCS consumption rate is faster than the formation rate. However, more tests are needed to verify this. Since all experimental results in Fan’s investigation were collected after the experiment, there was no instantaneous data during the test to tell how the deposition rate dropped.
3.4.3 Surface Areas of Cu-Si Alloys
The tests with two extremes of surface areas of 30wt%Si-Cu alloys were conducted to demonstrate whether various surface areas of alloys could affect Si deposition rate: 1. A regular cylindrical alloy (small surface area) 2. Very fine pellets (large surface area) Reaction conditions were similar in these two tests except the forms of the Si-Cu alloys. The details about the silicon rods and Cu-Si alloys for each test are listed in Table 3.5.
Table 3.5 Test conditions and measurements before the experiments for the different alloy surface areas [10] Test S1 Test S2 Half cylindrical Test Condition Fine pellets alloy Length (cm) 13.2 13.3 Si Starting Rod Average Diameter (mm) 4.98 5.02 Mass (g) 6.06 6.03 Average Thickness 9.6 N/A (mm) Length (mm) 94 N/A Cu-Si Alloy 0.589-1.65: 234g Pellet Diameter (mm) N/A 1.65-4: 69g Surface Area (cm2) 201 2017 Mass (g) 402 297 Experiment Duration (h) 27.5 27.5
19 Larger surface area should result in a higher silicon extraction rate. Even if extraction reaction is not the rate limiting step, the deposition rate should at least be the same. However, the results of the two tests tabulated in Table 3.6 show instead, a decrease in the deposition rate for the sample with the highest surface area. It was found that the Si pellets were sintered together after the experiments, which might decrease its actual surface area. EDX analysis on the sintered layer also revealed high silicon content, which may suggest that silicon back deposition occurred.
Table 3.6 Measurements after experiment for different Cu-Si surface areas Test S1 Test S2 Test condition Cylindrical alloy Pellet alloy Deposition Length 9.04 8.7 (cm) Diameter (mm) 12.38 11.34 Gained (mm) 7.4 6.3 Diameter growth 0.27 0.23 rate (mm/h) Si Rod Mass (g) 25.49 21.12 Gained 19.4 15.1 Deposition Rate 0.71 0.55 (g/h) Deposition 0.078 0.063 Rate/length (g/h/cm) Mass (g) 380.2 278.7 Cu-Si Alloy Loss 21.8 18.3
Another experiment done in Ebner (Linz, Austria) showed that a larger surface area has a positive impact on deposition rates. The reactor set-up is shown in Figure 3.8.
20
Figure 3.8 Experimental set-up in Ebner
The two tests utilizing 30wt%Si-Cu alloys but different stack ups, illustrated in Figure 3.9, were conducted to demonstrate the influence of larger surface areas on deposition rates.
Figure 3.9 Different alloys stacking up
21
Both reactions were performed for 20 hrs, and the reaction conditions were nearly the same, except the alloy weight in Test #1 that was 30% more than for Test #2. In Test #1 three alloy pucks were placed on the top of each other, and it was difficult for the gas flow to reach the gaps between the pucks. Several copper rings were added into the system in Test #2 to allow the gases to flow around the two pucks. Since the alloys could be exposed to more gas flow in Test #2, which in turn means there was more affective alloy surface area in Test #2 than that in Test #1, more TCS should be produced. The results showed that the Si deposition rate in Test #2 was 43% higher than in Test #1. Thus, this suggested that the TCS production rate might control the overall deposition rate. [20]
22 4. EXPERIMENTAL CONDITIONS AND PROCEDURE
The process set-up used in this investigation is very similar to Fan’s design in which two major reactions are involved:
Silicon Extraction: HSiCl3(g)+H2(g)=Si+3HCl(g) [4.1] Silicon Deposition: Si+3HCl(g)=HSiCl3(g)+H2(g) [4.2]
Both reactions occur in the same reactor but at different locations where the temperatures are significantly different. The advantage of having this kind of reactor is to cut costs of the process compared to the Siemens process by eliminating the fluidized bed reactor and distillation columns used in HSiCl3 production.
4.1 EXPERIMENTAL CONDITION DETERMINATION
In Fan’s and Tejedor’s investigations, the temperatures of silicon extraction reaction and silicon deposition reaction were controlled between 1100-1300oC and between 400- 600oC respectively, according to chemical thermodynamic analysis. To further confirm these conditions, a thermodynamic analysis was performed by utilizing HSC chemistry [21].
Figure 4.1 shows the equilibrium partial pressures of possible gas species formed from H, Cl and Si at different temperatures with excess solid silicon. The plot is used to determine which species are dominant at a certain temperature. The formation of copper gaseous species was also studies, and it was found that its amount was insignificant compare to o the silicon gaseous species. It can be seen that at temperatures below 1100 C, HSiCl3 o and SiCl4 are the major species while above 1100 C, SiCl2 starts to be dominant. It is o o interesting that the SiCl4 pressure at 1100 C is basically the same as at 400 C. This means that at those temperatures, there will be minimal reactions involving SiCl4 after the initial start up period.
23 Therefore, it appears that the most involved species is SiHCl3. It forms up to a pressure of almost 0.09 bars at 500oC, which at 1100oC it can only remain at 0,043 bars, meaning that roughly half of the SiHCl3 will decompose.
Figure 4.1 Partial pressure of possible gaseous species from H, Cl, and Si.
To determine optimum conditions we have to include all SiHxCly gas species. If we assume that we have thermodynamic equilibrium at both of the cold and hot reaction sites, then we can use this information to optimize the pressure condition. The sum of the equilibrium partial pressures of the various Si gaseous species is plotted in Figure 4.2.
24
1200oC 810oC
Si Gaseous Species Formation Favored
Figure 4.2 Sums of the equilibrium partial pressures of silicon gaseous species (HCl-H2-Si mol ratio is 1:1:50)
It can be seen that the lower the temperature, the more silicon can be extracted into the gas phase. The minimum total pressure is at about 1150-1200oC, which should be close to the optimum deposition temperature. At temperatures above 1200oC, the sum of partial pressures starts to increase so deposition is not favoured anymore. Therefore, thermodynamically speaking, the MG-Si should be kept at low temperature, like 300oC to allow more HSiCl3 formation. However, the actual temperature of the MG-Si was kept between 500-700oC to speed up the reaction kinetically. The silicon starting rod was kept between 1000-1100oC to favour Si deposition. Higher temperature would lead to higher energy losses as well as material issues.
Another consideration is the behavior of the various impurities. Some impurities can also form gaseous chlorides as discussed in Section 3. Would the formation of those impurity chlorides be favorable at the operation temperatures just decided above?
25 Chemical analysis showed that the major impurities in the metallurgical grade silicon used in the process are Fe, Al, Mn, Ti Cu, B and P shown in Table 4.1. The detailed analysis report is given in Appendix A.
Table 4.1 purity levels in MG-Si Elements ppmw Elements ppmw Elements ppmw Elements ppmw Si Major Ta <5 Ag <0.3 Sb <0.1 Fe 2800 Zn 4.5 Cd <0.3 K 0.072 Al 335 V 1.7 Te <0.3 S 0.069 Mn 55 Ge 1.6 Sr 0.23 As 0.065 Ti 35 Co 1.3 Pr 0.2 Pb <0.05 Cu 24 Ce 1.3 W 0.2 Bi <0.05 B 18 Nd 1.2 Nb 0.17 Th 0.039 P 16 Mo 0.71 Y 0.11 U 0.035 Cr 8.7 Mg 0.7 Na 0.1 Li 0.033 Ni 7.1 La 0.52 F <0.1 Zr 7 Cl 0.31 In <0.1 Ca 5.6 Se <0.3 Sn <0.1
The sums of the equilibrium partial pressure of Si, B, Al, Fe, P, Ti and Mn gas species are plotted in Figure 4.3. It is calculated for pure solid phases (Si, B, Al, Fe, P, Ti and Mn) for illustration purposes. The gaseous species of each element are listed in Appendix B. It clearly shows that with increasing temperature, the formations of Al, Fe, Mn, Ti and P gaseous species are more favored. Hence, those impurities should not deposit on the Si starting rod where the temperature is the hottest in the reactor. On the other hand, Boron may be a problem.
26
Figure 4.3 Sums of the equilibrium partial pressures of Al, Fe, B, Ti, Mn and P gaseous species
4.2 REACTOR DESIGN
The schematic drawing below illustrates the experimental set-up.
Figure 4.4 schematic of experimental set-up
4.2.1 Reactor
27 The silicon starting rod used is electronic grade silicon which will not introduce impurities to contaminate the deposited Si on the rod. The rod will be resistively heated to between 1000-1100oC by a DC power supply, which is shown in Figure 4.5. The radiation emitted from the rod will heat up and keep the Cu-Si alloy between 500-700oC. Attached water-cooling pipe to brass end caps and cooling air at the bottom of the quartz tube are used to prevent overheating. The temperature difference will produce HSiCl3 at the Cu-Si alloy side and transport it from the alloy side where the temperature is low, to the hot silicon filament by natural convection. Finally, the transported HSiCl3 can react with H2 to form solar grade polysilicon onto the hot starting silicon rod.
Figure 4.5 DC Power Supply
The reactor tube is made of quartz for three reasons: first quartz is transparent, through which observations can be done; second, quartz will not block the wavelength which emitted from resistively heated silicon rod, so the temperature of the rod can be monitored by optical pyrometer; and thirdly it is basically inert to HCl-H2 gases at those temperature. The dimension of the reactor is listed in Table 4.2
Table 4.2 Reactor Tube Dimension Outer Inner Volume of Length (cm) Diameter (cm) Diameter (cm) Reactor (cm3) Quart Tube 7.6 7.4 40.6 1746
28
The brass end caps are used to seal the reactor to make sure there are no gas leakages. One of three holes through the caps will allow the Cu rod connect to the silicon starting rod which can be resistively heated by passing current. The thermocouple and HCl & H2 gases inlet/outlet pipes go through the other two holes.
Chromel external heating elements are coiled outside of the quartz tube to heat the reactor above 700oC at which the silicon-starting rod starts to conduct current to be resistively heated.
4.2.2 Cu-Si Alloys
The investigation done by Fan [10] shows that hypereutectic Cu-Si alloy seems to have a higher silicon growth rate than a hypoeutectic alloy. Therefore Cu-Si alloys used in my experiments are all hypereutectic alloys. Cu and MG-Si were melted in an induction furnace. During melting, they were protected from air by blowing Ar gas above the charge. Once the alloy was molten, it was poured into high purity graphite moulds to form either puck or square form alloys shown in Figure 4.6. Cast alloys would solidify quickly under normal atmosphere and room temperature. Alloys with different compositions 30, 50 and 75wt%Si were made by the same method.
Figure 4.6 Cast Cu-Si alloys in puck and square shapes
29
Then the cast alloys were shaped to rectangles to have similar dimensions, in other words, similar reaction surface area, shown in Figure 4.7. Alloys were placed on Al2O3 dishes to allow for weight determinations as well as collection of any dust formed.
Figure 4.7 Alloys in rectangle shape
4.2.3 Silicon Starting Rod
The electronic grade silicon used in the process is roughly 5 mm in diameter shown in Figure 4.8. The lengths of the rod varied in different experiments. Chemical analysis results tabulated in Table 4.3 show impurities level in the ppb range (detailed report in Appendix A)
Figure 4.8 Electronic grade silicon starting rod
30 Table 4.3 Purity levels of silicon staring rods ANALYSIS ppmw ANALYSIS ppmw ANALYSIS ppmw Si Major Pb <0.05 W <0.01 Ta <5 Bi <0.05 Th <0.01 Cl 0.33 Fe 0.048 U <0.01 Ge <0.3 As <0.03 Cr 0.009 Se <0.3 K 0.021 Na 0.007 Ag <0.3 Cu 0.018 Ni 0.005 Cd <0.3 B 0.016 S <0.005 Te <0.3 Li <0.01 Al 0.004 F <0.1 P <0.01 Mg 0.003 In <0.1 Ca <0.01 Mn 0.003 Sn <0.1 Zr <0.01 Ti <0.001 Sb <0.1 Nb <0.01 V <0.001 Zn <0.05 Mo <0.01 Co <0.001
An actual reactor set-up picture is shown in Figure 4.9.
6 7
4 24 3 3 1 5
Figure 4.9 Actual reactor set-up
Where, 1. Starting Cu-Si alloy rectangle 2. Silicon starting rod 3. Graphite connector
31 4. Cu rod which connected with wires to the DC power supply
5. Al2O3 dish 6. Chromel external heating element 7. Quartz tube
4.2.4 Power Supply
Two power supplies were used in the experiments: