Removal of Dissolved Al and Ca in Si by SiO2 Additions and Mechanical Stirring

Transcription

1 DEGREE PROJECT, IN MATERIALS SCIENCE, SECOND LEVEL STOCKHOLM, SWEDEN 2015 Removal of Dissolved Al and Ca in Si by SiO2 Additions and Mechanical Stirring MIKAEL SANDELL KTH ROYAL INSTITUTE OF TECHNOLOGY INDUSTRIAL TECHNOLOGY AND MANAGEMENT

2 Abstract In the oxidative refining of metallurgical grade silicon the loss of Si to the slag in the form of SiO2 is an economical concern. The purpose of this report is to investigate the possibility of using SiO2 and mechanical stirring to remove Ca and Al as a substitute to the oxidative refining. In the experiments graphite crucibles were used in a vertical resistance furnace and controlled argon atmosphere. The removal rate of Ca and Al are measured by X-ray fluorescence and the slag is examined in a scanning electron microscope. The slag formation kinetics are examined and a calculation of the activities of Ca, Al and their respective oxides in the slag phase is conducted. The driving forces of creating CaO and Al2O3 in this system is calculated to better understand the behavior of the Ca and Al removal. The results show that removal of dissolved Ca and Al by mechanical stirring is possible and in this setup a stirring time of 20 minutes is sufficient since no more refining can be obtained by increasing it.

3 Acknowledgements I would like to express my appreciation to all the people at Elkem Technology for giving me the opportunity to work with this project. I am very grateful for the support of my supervisors Dr. Björn Glaser, for all the effort he has made in teaching me how to work in a laboratory environment, Dr. Kjetil Hildal and Dr. Jesse White for going out of their way to search for a project for me at Elkem and helping me sort out the proceedings of this project and coordinating my work with the laboratory in Norway. They also helped me with the experimental planning and provided information about what has been done before. Lastly, I would like to express my gratitude toward my fellow students and co-workers at the Royal Institute of Technology for all the help with my calculations and for all the helpful advice.

4 Contents 1. Introduction MG-Si Production process SOG-Si Production Process Acid Leaching Oxidative Refining Vacuum Treatment Solidification Refining Mechanical Stirring Thermodynamic Data Silicon Melt Slag System Experimental Materials Experimental Setup Experimental Procedure Experimental Conditions Analysis Si Analysis Slag Analysis Results & Discussion Compositional Results Slag Formation Slag Analysis Slag Composition Activity Calculation Ca and Al Activity Oxide Activity Calculation of the Driving Force Sources of Error Conclusions Future Work References... 21

5 1. Introduction Silicon can be used in a wide array of different applications. As of 2010 the automotive industry made up about 50% of the global demand and photovoltaic cells (PV) and semiconductors around 10% [1], [2]. The PV industry is expected to increase since solar energy is infinite and clean. Solar grade silicon (SOG-Si) which is used for PV cells is obtained by removing impurities from metallurgical grade silicon (MG-Si). 1.1 MG-Si Production process The raw material for MG-Si production, SiO2, is abundant in nature in the form of quartz and sand. The production of MG-Si is done by reducing the silica with carbon in an electric arc furnace, also known as carbothermic reduction. The silicon is reduced according to the following total reaction: SiO2(s)+ 2C(s) Si(l)+ 2CO(g) (1.1) The resulting silicon has a purity of around 98% with common impurity elements such as aluminium, iron and calcium [1]. A general overview of the production process is shown in figure 1. Figure 1. Overview of the silicon production from the basic materials SiO 2 and C. [3] The charge material consists of quartz sand, carbon and electrical energy. Electrical energy is fed through the electrodes at the center of the furnace. When the charge material has been heated and the reactions are complete the liquid Si is drained at the bottom and subsequently 1

6 solidified and crushed. Even though the Si is drained at the bottom it is important to note that this process is continuous and the furnace is never emptied as long as it is active. A downside of this process is that it produces an exhaust gas that contains a lot of SiO that needs to be recycled into the process in order for the silicon yield to be high enough. The exhaust from the furnace also contains large amounts of SiO SOG-Si Production Process When producing SOG-Si the quality of the silicon obtained through the electric arc furnace is too low. The impurities decrease the electricity generation in the solar cells [4] and need to be removed. Further refining of the MG-Si can be done in many ways and has been investigated in previous articles [5] [6]. The most common methods to remove impurities involve acid leaching, reactive gas blowing with O2, Cl2 or other gases, vacuum treatment and finally solidification refining. The order of these procedures are shown in figure 2. Figure 2 Flow schedule of the production from MG-Si to SOG-Si Acid Leaching Silicon has a high resistance against corrosive substances. Acid leaching is, because of this property, a suitable method to remove impurities that would be hard to remove by other methods. Iron, being harder to oxidize than silicon, would be such an element. This method is also suitable for titanium removal. Although the oxidation of titanium is easier than the oxidation of silicon, measurements by Morita and Miki [6] show that the activity coefficient of titanium inside silicon is too low to make oxidation of titanium sufficient. Therefore acid leaching is a suitable replacement. 2

7 1.2.2 Oxidative Refining The most common method for reactive gas blowing involves oxidative refining in which oxygen is blown through a silicon melt in order to oxidize the impurities and separate them into a slag system. This method is suitable for all the elements that have a lower Gibbs energy of oxidation than silicon. These elements, Ca, Al and Mg, also have relatively large vapor pressures [6] so removal by either oxidation or vacuum treatment is suitable. The oxidative refining that takes place when refining MG-Si creates a lot of unnecessary oxidation of the silicon. It is a good method for removal of these impurities but the resulting slag contains large amounts of SiO2. It is known that one can use SiO2 to remove impurities but that it is less efficient [1]. The idea behind this report is to investigate the practical possibility of using SiO2 to remove the impurities by mechanical stirring. The removal by SiO2 is governed by these two reactions: 4Al( l) in Si 3SiO 2( s) Si( l) 2Al2O3 ( s) (1.2) 2Ca( l) in Si SiO2 ( s) Si( l) CaO( s) (1.3) The economic incentive is to increase the refining efficiency by decreasing the amount of Si that is lost to the slag phase. When using SiO2 the only byproduct, in theory, is pure Si. The benefit of employing oxygen blowing compared to using SiO2 when removing impurities is that heat is generated by the reactions with oxygen and that the melt is simultaneously stirred. When adding SiO2 not as much heat is produced by the reactions with the impurities so the majority of the heat has to be provided externally. It is also necessary to manipulate the kinetics in another way. This report is an investigative trial to determine the possibility of employing mechanical stirring along with addition of SiO2 in order to remove Ca and Al impurities. If the trials are successful they will lead to an up scaled furnace trial at Elkem Technology s facilities at Kristiansand Vacuum Treatment Exposing the silicon melt to a vacuum atmosphere will drive all the elements with a lower vapor pressure than silicon out of the system. It has been examined in a previous article [7] that it is a good method for phosphorus removal. As mentioned before it is also appropriate for removal of Al, Ca and Mg Solidification Refining This method utilizes that most impurities in silicon have low solubility limits. If the solidification front of the silicon is planar the impurities will remain in the melt that is left on top of the solidified surface. When the silicon is completely solidified the upper part of the piece, which contains the most impurities, is removed, leaving the purer silicon for further refining. For the elements that this method doesn t work for, mainly phosphorus and boron, other more expensive methods have to be used. 1.3 Mechanical Stirring The main topic of this article is refining of silicon by employing mechanical stirring. When performing the oxidative refining of silicon it is a natural benefit from the gas blowing that the system is stirred simultaneously. An alternative way of achieving the same effect is to employ 3

8 mechanical stirring with an impeller blade. The literature concerning this topic is scarce, but a few trials have been successfully performed in a previous study [8] in which experiments were carried out to investigate the separation of slag from silicon in a Si-slag system. The experiments indicated that the mass transfer of impurities to the slag is very fast when employing mechanical stirring compared to conventional gas blowing. When conducting similar experiments without any mechanical stirring the reaction between the silicon and the slag will be very slow since the kinetics of the system is not manipulated in any way. The reaction rate is increased when employing mechanical stirring and when the reaction is close to equilibrium the silicon tended to separate from the slag. This is explained with reaction (1.4). G G A W Reaction (1.4) The driving force, ΔG, is controlled by these forces; the driving force for the reaction to occur, ΔGReaction, the interfacial tension between the two phases that form, σa, and the applied work ΔW, in this case the mechanical stirring. In White s article it was shown that, for the experiments with the mechanical stirring, the content of impurities in the metal phase would appear to be closer and closer to a steady composition after a certain time. This indicates that the driving force for the reaction gets closer and closer to zero, i.e. equilibrium. As the ΔGReaction gets smaller and smaller the contribution from the interfacial tension becomes of greater importance. This means that the system tends to separate the two phases, i.e. slag and metal, in order to reach the lowest possible energy state for this system. The interfacial tension between the two phases becomes the dominant force and the system strives to separate the silicon and the slag phase [8]. For the experiments conducted later in this report it means that all the phases in the system will, after being provided sufficient stirring, separate from each other. 1.4 Thermodynamic Data Silicon Melt The data regarding interaction and activity coefficients between the elements in the silicon melt have been examined in previous reports [4] [6]. Since the silicon s main contaminants are Ca and Al, see table 1, and the content of the other elements are low in comparison, the interaction coefficients for these elements are the only ones that are considered. Table 1 Composition of the silicon used in the experiment. Element Ca Al Fe Mg Composition [mass%] The thermodynamic data that was used for the later calculations are listed in Table 2. 4

9 Table 2. List of thermodynamic properties obtained from the literature. ε Ca Ca = T References (1.5) [4] [6] ε Ca Al = 6.46 (1.6) [4] ε Al Al = 105 T 40.1 (1.7) [4] [6] 0 ln(γ Ca(l)in Si ) = T (1.8) [4] [6] 0 ln(γ Al(l)in Si ) = 3610 T (1.9) [4] [6] These equations will be used to calculate the activity of Ca and Al in the melt and how it changes with time and the stirring rate Slag System The slag system that is formed when Ca and Al is removed will consist mainly of SiO2. Reaction (1.2) and (1.3) are the reactions that take place in the system and the standard Gibbs free energies of these reactions, and their partial reactions, are shown in Table 3. Table 3. List of reactions and their respective standard Gibbs free energies. Reaction Gibbs Energy at 1550 ºC [J/mol] Reference Si(s)+O 2 (g)=sio 2(s) [10] Al( l) 3O ( g) 2Al O ( s) [10] 4 in Si Ca( l) in Si O2 ( g) 2CaO( s ) [10] Si( s) Si( l) [10] Si( l) O2 ( g) SiO2 ( s) [10] Al( l) 3SiO ( s) 3Si( l) 2Al O ( s) in Si Ca( l) in Si SiO2 ( s) Si( l) 2CaO( s) The ternary phase diagram for the SiO2-Al2O3-CaO slag system has been modelled in a previous study [9] and at the temperatures relevant for the experiments conducted in this report separate calculations for the phase diagrams were performed in ThermoCalc and compared to the literature. The phase diagrams were used to determine the amount of SiO2 that needs to be used for the experiments. The phase diagram of the SiO2-CaO-Al2O3 system that was calculated in ThermoCalc can be seen in Figure 3. 5

10 Figure 3 Isothermal section of the ternary SiO 2-CaO-Al 2O 3 phase diagram. The phase diagram will be used to investigate whether the calculated slag composition will be liquid or a multi-phase system. If the added SiO2 amount yields a multi-phase slag the amount has to be changed. Compared to the phase diagram in Mao s work it is a good match [9]. 2. Experimental The general procedure of the experiments is that SiO2-sand will be added to a Si-melt containing trace amounts of Ca and Al. The two most interesting topics is to examine how efficient the removal of these dissolved elements will be and how fast the slag will be formed. 2.1 Materials The material that is used was prepared by Elkem Technology in Norway. The MG-Si was prepared as large metallic pieces and the SiO2 was delivered as a coarse sand. The analysis of these can be seen in Table 4. Table 4. Composition of the raw materials used in the experiments. Substance SiO 2-sand [wt%] MG-Si [wt%] Si Al Ca Fe Mg SiO CaO Al 2O

11 For each run around 80 g of MG-Si was used. The silicon pieces that were too big to fit inside the crucible was crushed into smaller pieces. The SiO2 sand was sieved so that only one particular grain size would be used. The sand size range used in these experiments was between 1 and 1.4 mm in size. The silicon is inserted into the graphite crucible and the SiO2 sand is wrapped around a graphite impeller using an organic plastic foil, see Figure 4. Figure 4. The SiO 2 sand was wrapped around the graphite impeller with an organic plastic foil. When the SiO2 reacts with the dissolved Al and Ca it will form a slag. It is important to keep the slag in a liquid phase in order to improve the kinetic conditions of the reactions. A variety of assumptions are made in order to determine the appropriate amount of SiO2 that needs to be added to keep the slag liquid. These assumptions are only a starting point and the calculated composition will have to be compared to the slag systems phase diagram in order to determine what slag phase this will yield. The assumptions are listed below: - The Al and Ca compositions are decreased to 0.1 mass% from their original values. - All the Ca and Al will react with SiO2 to form their respective slag component, they will not react with anything else. - The other elements, Fe and Mg, is not considered to be of importance in the calculations since the amount is low (60 and 7 ppm respectively). - The SiO2 that is not reacted with Al or Ca will make up 50% of the slag that is formed. From these assumptions it is a simple step to figure out how much CaO and Al2O3 is formed and, subsequently, how much oxygen is needed from the SiO2. The amount of SiO2 needed to remove this amount of Al and Ca from the Si is equal to 2.1 g per 80 g of Si sample. However, in order to keep the SiO2 amount in the slag to 50 mass% the new value becomes 7

12 4.82 g. This extra amount is simply equal to the amount of CaO and Al2O3 that is formed in grams. Table 5 shows what the resulting slag composition will be. Table 5. The initially calculated slag composition. Slag Composition SiO2 [mass%] 50 CaO [mass%] Al2O3 [mass%] When examining this composition in the phase diagram, Figure 5, it is shown that the slag is in a two phase region and not completely melted. Figure 5 Composition of the slag that is formed during the experiments. This composition can only be manipulated by increased additions of SiO2, provided that the Al and Ca removal isn t changed in a dramatic way. To ensure that the slag is liquid an additional amount of SiO2 will be added, a sufficient amount was calculated to be 6 grams, making the SiO2 content in the slag around 60%. 8

13 2.2 Experimental Setup The furnace that was used for the experiments is a vertical resistance furnace and the schematics of it can be seen in Figure 6. Engine for stirring Gas Outlet Gas Inlet Steel Rod Quenching Chamber Alumina Tube Crucible Holder Crucible with impeller inside Thermocouple Gas Inlet Figure 6. Schematic cross section of the furnace used for the experiments. The furnace is divided into two sections. The top part is the quenching chamber and the bottom part is the furnace chamber. The quenching chamber is made of brass and, along with the bottom part with the gas inlet, it is water cooled throughout each experiment. At the bottom of the steel rod there is a graphite extension where the crucible holder can be attached. This extension is there to protect the steel rod from the hottest zone in the furnace. The steel rod contains an inner rod where an impeller can be fastened. The steel rod is held in place by a lifting device. A separate platform, which is not connected to the lifting device, can be lowered and fastened above the steel rod and an engine can be connected to the inner rod in order to enable rotation of the inner rod and stirring of the sample. The thermocouple is a type B thermocouple for measuring temperatures up to 2073 K. The crucible and impeller designs can be seen in Figure 7. 9

14 O 38 M8 threaded hole R Figure 7. Impeller and crucible designs. The measurements are in mm. The crucible is manufactured by drilling four holes in a cylindrical graphite piece. The baffles that are consequently produced are there to avoid bulk rotation and vortex formation in the melt. 2.3 Experimental Procedure The sample is put into a larger graphite crucible holder. This holder is connected to the graphite extension on the steel rod with two small graphite screws. After the sample is connected to the rod it is lowered in to the quenching chamber and the furnace is sealed. It is evacuated using a vacuum pump for 30 minutes. Afterwards the furnace is flushed with argon gas for 30 minutes and the experiment can be initiated. Due to the reaction tubes sensitivity to thermal shock the temperature can only be increased by a maximum of 3 C per minute. For the same reason the sample is lowered in 3-4 cm intervals every two minutes until it reaches the hot zone. The time it takes to reach the hot zone varies therefore between each run but it is usually around 20 minutes. The thermocouple located at the bottom of the furnace is placed around 1 cm below the bottom of the sample holder. This ensures accurate temperature measurement of the melt. Inside the rod there is a secondary rod in which an impeller can be fastened. The position of this impeller is controlled manually and to avoid melting of the plastic foil the impeller is kept at a high position when the sample is lowered into the furnace. When the sample has reached the hot zone it is kept there for 10 minutes before the impeller is submerged into the melt to be sure that the silicon has been melted. The impeller is then rapidly lowered into the melt to ensure that all the SiO2-sand that is wrapped around the impeller will end up inside the melt instead of on top of it. The zero time 10

15 for the experiments is defined as the time when the stirring of the impeller has started. The conditions for each experiment is shown in the next section. When the experiment is finished the sample is rapidly elevated into the quenching chamber. Argon gas is flowing at a high rate at the inlets directly onto the sample holder. It is quenched for 30 minutes before the furnace is opened and the sample is retrieved. 2.4 Experimental Conditions The stirring time and stirring rate of the experiments are shown in Table 6. Table 6 Displaying time and stirring rate conditions for each experiment during the slag formation kinetics investigation. Stirring rate [rpm] Times [min] The temperature for the experiments was always 1823 K. The reason for this is simply that it would require too many experiments to determine a temperature dependence of the Ca and Al removal. Also 1823 K is in the range of what is commonly used for MG-Si refining. The stirring rates were decreased to two different rates in order to lower the amount of experiments that would be performed. A higher stirring rate than 200 rpm is considered to be too high for the furnace that is used since the stirring rod becomes too unstable. The gas flows in the various steps are not controlled via a flow meter but is only controlled via a set of gas bubblers connected between the furnace and the gas tube. 2.5 Analysis Si Analysis The compositional analysis of the Si is done by X-ray fluorescence (XRF). Figure 8 shows where the samples that are sent away for analysis are taken from the crucible. It is not taken from the same place in each sample since the area where the slag is concentrated differs between the samples. The piece that is removed from the crucible is where the smallest amount of slag is visible. The pieces are cut off using a carbide blade saw. The graphite crucible walls and any slag residuals are removed by grinding it off. The method of analysis is sensitive to small amounts of contaminations so to make sure that the slag residuals are completely removed a Dremel was used to grind off the small slag pieces that were found after the samples were examined in a magnifying glass. 11

16 Figure 8. Section of the sample that is cut away and sent for analysis Slag Analysis The slag that is formed during the experiments is too small to be able to send away for XRF analysis so it is measured in a scanning electron microscope (SEM) for a semi-quantitative energy-dispersive X-ray spectroscopy (EDS) analysis. Because of the small slag amount in each sample it was difficult to be consistent when selecting where to analyze the slag. The area where the slag was analyzed was therefore where the slag was concentrated, which was different between each sample. The slag that is selected in each sample is cut out and put into a poly resin sample holder for SEM analysis. The sample is then polished on SiC grinding plates and then further polished on a cloth piece with a liquid diamond crystal suspension. Lastly the samples are coated with a gold layer to make an SEM analysis possible. The experiments where the slag was analyzed were the ones with 5 and 20 minutes stirring for both stirring speeds. The reason for this is explained in the results & discussion section. 12

17 Element Content [mass%] Element Content [mass%] 3. Results & Discussion 3.1 Compositional Results The compositional results from the experiment will be presented as a function of the stirring time. Two separate stirring speeds were used in the experiments so the results will be presented as two separate graphs Rpm Ca Al Stirring Time [min] Figure 9. Content of both Ca and Al in the Si as a function of the stirring time for the experiments run at 100 rpm. In the samples that were run at 100 rpm the removal of Ca is a little bit faster than the removal of Al and after 20 min stirring there is little or no additional removal of either elements, indicating that an equilibrium has been reached. 200 Rpm Ca Al Stirring Time [min] Figure 10. Al and Ca content in the Si as a function of the stirring time for the samples that were run at 200 rpm. 13

18 The same tendencies are shown for the samples that were run at 200 rpm. A comparison between the two graphs reveal that the removal of Ca is a little bit faster in the 200 rpm samples but the removal of Al is a bit slower. The reason for this is hard to examine since the difference is so low but it could be an effect of the increase in reaction kinetics due to the increased rotation speed. To examine why Ca is removed faster than Al the equilibrium state for the two separate reactions between the dissolved elements and SiO2 are examined. Al(l) in Si 3/4SiO 2(s) 3/ 4Si(l) 1/2Al 2O3(s) (3.1) Ca(l) in Si 1/ 2SiO 2( s ) 1/ 2Si(l) CaO(s) (3.2) K (3.3) 3.1 K (3.4) 3.2 The equilibrium constants are calculated with the standard Gibbs free energies for the two reactions, 1.2 and 1.3, which are presented in the introduction part. The Gibbs energies that are calculated will have to be revised so that only one Ca and Al atom is considered to avoid problems when comparing the two equilibrium constants. This is done by dividing the Gibbs energy for reaction 1.2 by 4 and reaction 1.3 by 2. The expression for the equilibrium constant is well known as: K G exp (3.5) RT These values show a strong tendency towards creating both Al2O3 and CaO when introducing SiO2. The equilibrium constant for the creation of CaO is larger than the constant for Al2O3 formation so thermodynamically the formation of CaO should lead to that the Ca content in the Si should be lower than the Al content when an equilibrium has been reached. However, when trying to understand why Ca is removed faster than Al there are more things to consider than the thermodynamics. The reaction rate is determined by the Gibbs energy for each reaction, i.e. the thermodynamics, and the driving force for creating each reaction product. In this case the solubility limit for CaO in the CaO-Al2O3-SiO2 slag system is around 55 mass% while the limit for Al2O3 is around 30 mass%. This creates a larger driving force for creating CaO to enter the slag system. This might explain why Ca is removed faster than Al. Further understanding requires an analysis of the slag. The analysis of the oxide activity in the slag as well as the Ca and Al activity will be presented later in the report and a comparison between the equilibrium state and the current situation is made in order to determine the driving force for the respective reactions. 3.2 Slag Formation It takes some time to form the slag in the experiments and both the speed of this slag formation and the dissolution of the SiO2 into the slag is examined. 14

19 Figure 11. An overview of the samples run at 100 rpm. The samples (a), (b), (c) and (d) were run for 5, 10, 20 and 30 minutes respectively. After 5 minutes it can be seen that the slag has not yet been completely formed, most of the SiO2 sand that was added is still visible as grains. In the (b) sample some of the grains are still visible. In sample (c) the slag is clearly visible in the upper part of the sample. However, there are still some visible white grains in the slag that is believed to be unreacted SiO2. There is a small amount of slag in the sample that was stirred for 30 min but the small amount that is visible indicates that there is little change in the slag after 20 minutes, as is also indicated by the compositional results. Since the SiO2 grains are still visible in the slag this might indicate that the amount of SiO2 amount that was added is too high and it is not possible for this amount to dissolve completely within this time frame. 3.3 Slag Analysis As mentioned in the experimental part the only samples that were used in the SEM analysis of the slag were the samples that were stirred for 5 and 20 minutes. The reason for this is that there was limited time to analyze the composition and, as indicated by the compositional results, there is practically no further removal of Al and Ca after 20 minutes so there was no need to analyze the samples that were stirred for 30 minutes Slag Composition The results of the slag analysis are shown in Table 7. It should be mentioned that the EDS analysis of the slag is semi-quantitative so for an accurate calculation it is not an appropriate method. However, a general idea of the slag composition and compositional tendencies can be obtained if no other alternative is available. Table 7. Average composition of the slag samples that were analyzed. Sample SiO2 [wt%] CaO [wt%] Al2O3 [wt%] Sample Condition A rpm, 5 min B rpm, 20 min C rpm, 5 min D rpm, 20 min 15

20 The slag was analyzed with an area analysis at a minimum of 8 areas in the slag of varying sizes and an average composition was calculated for each sample. The results that were expected was that both the Al2O3 and CaO content in the slag would increase slightly between the 5 and 20 minute samples since this is how the composition in the Si phase is changed. These values don t fully correlate with this expectation. The SiO2 composition is increased after 20 minutes, probably due to SiO2 dissolution into the slag, and this might explain why the CaO and Al2O3 values don t increase as they should. Another factor that affects the analysis is where the slag is analyzed. Figure 12. SEM mapping of the interface between the silicon and slag phase for the A sample. In Figure 12 it can be seen that the SiO2 grains are still in the slag. Depending on where the slag analysis is made, i.e. how close it is to the SiO2 grains, the SiO2 composition will vary. Another aspect of the analysis is that only one sample per experiment was analyzed and if the slag composition is not evenly distributed the composition can vary greatly. 3.4 Activity Calculation The activity of the dissolved components in the Si as well as the activity of the oxides in the slag have been calculated. For the slag system FactSage was used and the composition of the slag system, measured in the SEM for each sample, was entered into the software. These activities were then used to compare with the equilibrium state in order to further explain why CaO is formed much quicker than Al2O3. The reference state that is used for the respective components is pure liquid for the metals; Ca, Al and Si, and pure solid for the oxide elements. 16

21 Activity of Ca Activity of Al Ca and Al Activity The activities of these components were calculated using the following equations. ln γ ln γ ε x ε x (3.5) a x X x x X x x x y x y x (3.6) Where γ equals the activity coefficient at infinite dilution, the respective ε-values represent the interaction parameters between the elements themselves, i.e. between Ca-Ca and Al-Al, and the interaction coefficients between Ca and Al. γ represents the real activity coefficient. The x represents the molar fraction of each element and ɑ is the activity of the examined element. Equation 3.5 was used to find the activity coefficient for the respective elements at each compositional value that was calculated. When this value was calculated it was entered into equation 3.6 to find the activity of the component. In the materials section of the experimental part the composition of the Si show small traces of Mg and Fe, but since the amount is so small the effect of these elements on the activity of Ca and Al have not been considered. The respective values for each parameter in the equation has been taken from the literature and were presented in the introduction section. Activity 100 Rpm 1.00E E E E E E E E E-06 Ca Al Stirring Time [min] Figure 13. Activity of Ca and Al in the experiments run at 100 rpm. 17

22 Activity of Ca Activity of Al Activity 200 Rpm 1.00E E E E E E E E E Stirring Time [min] Ca Al Figure 14. Activity of Ca and Al in the experiments run at 200 rpm. The activity graphs are very similar to each other in the same way that the compositional graphs were also similar. One noticeable factor is that the activity of Ca is considerably lower than the Al activity Oxide Activity At the initial stages no Al2O3 or CaO is present, i.e. the activity is 0 for these components, so the introduction of SiO2 into the silicon melt will create a driving force to create these components. The composition of the slag analysis for the 5 minute samples was entered into FactSage to obtain the activity of the respective components. The results of the Factsage calculation is shown in Table 8. Table 8. Calculated activities of the slag components at 5 minutes for two separate stirring speeds. Component Stirring Speed Activity CaO Al2O CaO Al2O Calculation of the Driving Force These results are used to calculate the Q-value for the respective reactions to form CaO and Al2O3 to calculate the driving force towards creating the two components. Only the 5 minute samples were used in the activity calculation. The reason for this is that these samples are the furthest away from an equilibrium so the driving forces to create each slag component should be largest. The equations 3.7 and 3.8 show the equation that was used to calculate the driving force. 18

23 Q 3.1 a a 1/2 Al 3/4 a G G 2 O3 (s) Si(l) exp (3.7) (s) aal(l) RT 3/4 SiO2 Q 3.2 a a CaO(s) 1/2 SiO2 (s) a a 1/2 Si(l) Ca(l) G G exp RT (3.8) The Q-value is the actual state of the reaction and the activity of the Si that is formed is set to 1 since the majority of the metal phase is Si and the SiO2 activity is also set to 1 since there are still visible SiO2 grains in the slag. The only remaining activities are the activities of the dissolved elements and the slag components. For the sample run at 100 rpm the following driving forces are calculated for CaO and Al2O3 formation. The calculated values are in J/mol. G CaO a CaO( s) ln RT G ln CaO 6 aca( l) (3.9) G Al 1/ 2 a 1/ 2 Al O3 ( s) ln ln ( ) O RT GAl O a Al l (3.10) Similarly the driving forces for the sample run at 200 rpm was calculated as: G CaO RT ln G CaO (3.11) / ln G Al O RT GAl O (3.12) There is a large difference in the driving force between the two reactions at 100 rpm, roughly 3.5 times larger for CaO formation. At 200 rpm the difference in driving force is even larger which explains why the Ca content is lowered faster. 3.5 Sources of Error The experiments had some slight issues concerning the crucible material. Some of the crucibles cracked during the runs and the silicon poured out into the crucible holder. This was later found to depend on that some of the crucibles were of poor quality and more porous than the other crucibles. If the same problem exists, although in smaller scale, in the successful experiments it means that a greater amount of Si forms SiC in the crucible walls. How this affects the Ca and Al removal was not investigated. The method of inserting the SiO2 yields problems with experimental consistency. Since the impeller is inserted by manual force it is difficult to maintain the same insertion speed. If it is too slow the plastic foil will melt before the impeller reaches the melt surface and the majority 19

24 of the sand will rest on the melt surface instead of being put inside the melt along the impeller. If this had any major effect on the removal of Al and Ca is difficult to determine. The plastic foil is assumed to burn away but some of the carbon content in the plastic might be inserted into the melt yielding unwanted reactions between the dissolved elements and the carbon. The slag analysis was performed in an SEM which is not very accurate for analyzing elements which are not conductive. Additionally the samples were coated with a gold layer which further interrupts the analysis of the slag elements. The resulting slag analysis might not be representative of the slag. Furthermore this means that the activity calculation in FactSage is also inaccurate. The thermodynamic data for this system is not very well evaluated, i.e. the interaction parameters and the activity coefficients might be inaccurate, and the calculation of the activity of the dissolved Al and Ca might be incorrect. The slag that was analyzed was the slag that was still inside the sample. Since the impeller was removed in the majority of the samples prior to quenching some of the slag is removed from the sample. If this slag has another composition than the one remaining in the crucible was not analyzed. Furthermore, if the composition of the slag on the impeller is different than the slag in the crucible it might also mean that the composition of the slag in the sample is not evenly distributed. 4. Conclusions In this report the possibility of removing dissolved Ca and Al in Si by using mechanical stirring has been examined. Aside from this the kinetics of the slag formation have been examined and a brief calculation of the activities and driving forces of the components have also been conducted. What can be concluded is listed below. - The removal of Ca and Al is possible by using mechanical stirring. Ca was removed faster than Al and the lowest values of Ca and Al that were obtained were and 0.11 mass% respectively. - A stirring time longer than 20 minutes in this particular setup will probably not be necessary since little or no further refining occurs beyond this point. - There was no particular difference between the two different stirring speeds, 100 rpm should be sufficient. - After 20 minutes a slag is formed although SiO2 grains are still unreacted which might indicate that too much SiO2 sand was used. - The driving force for creating CaO is larger than Al2O3 so the Ca content will decrease faster. 4.1 Future Work In the activity calculations the values for the interaction parameters and activity coefficients are inaccurate and the calculations should be remade when these values are updated. 20

25 The slag analysis had to be made in an SEM due to the small amount available in the samples. If these experiments are carried out in a larger scale a proper slag analysis will yield more accurate results. The method of adding the SiO2 sand into the melt as well as the stirring conditions could be improved in a larger scale furnace. As mentioned in the experimental part the insertion of the stirring impeller in the current setup is manually controlled so a computer controlled setups is preferable. References [1] M. Andersson, T. Sjökvist and P. Jönsson, Processmetallurgins Grunder, Stockholm: Institutionen för Materialvetenskap, KTH Stockholm, [2] E. University, "High Si-alloys: Market and customer requirements - Powerpoint Presentation," [3] NTNU, "Norges teknisk-naturvitenskapelige universitet," [Online]. Available: [Accessed ]. [4] M. Takahiro, M. Kazuki and S. Nobuo, "Thermodynamic Properties of Si-Al, -Ca, -Mg Binary and Si-Ca-Al, -Ti, -Fe Ternary Alloys," Materials Transaction, vol. 40, no. 10, pp , [5] B. Bathey and M. Cretella, "Review Solar-grade silicon," Journal of Materials Science, vol. 17, pp , [6] K. Morita and T. Miki, "Thermodynamics of solar-grade-silicon refining," Intermetallics, vol. 11, pp , [7] J. Safarian and M. Tangstad, "Kinetics and Mechanism of Phosphorus Removal from Silicon in Vacuum Induction Refining," High Temp. Materials Processes, vol. 31, pp , [8] J. White, "Equilibrium and Kinetic Considerations in Refining of Silicon," Royal Institute of Technology, Department of Materials Science and Engineering, Stockholm, [9] H. Mao, M. Hillert, M. Selleby and B. Sundman, "Thermodynamic Assessment of the CaO- Al2O3-SiO2 System," The American Ceramic Society, vol. 89, pp , [10] E. T. Turkdogan, Physical Chemistry of High Temperature Technology, New York: Academic Press,

26