中文   |   English   |   Email Alert

Boron Removal Method, Technology and Process for Producing Solar Grade Silicon by Metallurgical Method

Yi Zeng , Yongsheng Ren , Wenhui Ma , Hui Chen , Shu Zhan , Jing Cao

Progress in Chemistry ›› 2022, Vol. 34 ›› Issue (4) : 926 -949.

PDF
Progress in Chemistry ›› 2022, Vol. 34 ›› Issue (4) : 926-949 DOI: 10.7536/PC210710
research-article
Scan to read & Share
Boron Removal Method, Technology and Process for Producing Solar Grade Silicon by Metallurgical Method
Author information +
History +
PDF

Abstract

As a kind of green renewable energy, solar energy has attracted wide attention, and impurity removal is a necessary purification process to obtain solar grade silicon from metallurgical grade silicon, which is very important for the preparation of silicon-based solar cells. The new technology for preparing solar grade polysilicon by the metallurgical method has become the focus of research and development because of its advantages such as low energy consumption, low cost and less pollution. However, the effective removal of boron is one of the most severe challenges we face. In this paper, the thermodynamic and kinetic properties of boron (solubility, diffusivity, diffusion coefficient, mass transfer coefficients and activity coefficient) and the research topics of boron removal in recent years (gas blowing, slag treatment, plasma treatment, acid leaching and solvent refining) are reviewed. It is found that solvent refining is a promising method to obtain high purity silicon. The enrichment rate of silicon and the removal rate of boron can reach more than 90%. Additives can strengthen the formation and precipitation of borides to improve the boron removal process, and the subsequent almost can be completely eliminated, which will not cause pollution to the refined silicon, which will be more effective in boron removal and increase the practicability of the process. At the end of the paper, several deboration processes are compared and analyzed, and the application prospect of metallurgical process is forecasted.

Contents

1 Introduction

2 Properties of boron in silicon

2.1 Solubility of boron in silicon

2.2 Diffusivity of boron in silicon

2.3 Diffusion coefficients and mass transfer coefficients of boron in silicon and slags

2.4 Activity coefficients of boron in silicon and slags

2.5 Segregation coefficients of boron between solid silicon and solvent

2.6 The technical difficulties and challenges of boron removal in silicon

3 Boron removal process

3.1 Boron removal by gas blowing

3.2 Boron removal by slag treatment

3.3 Boron removal by a united refining technique combined gas blowing with slag treatment

3.4 Boron removal by plasma refining

3.5 Boron removal by acid leaching

3.6 Boron removal by solvent refining

3.7 Application of boron removal technology by metallurgical method

4 Conclusion and prospect

Graphical abstract

Keywords

solar energy / metallurgical grade silicon / boron removal / metallurgical process / silicon purification

Cite this article

Download citation ▾
Yi Zeng, Yongsheng Ren, Wenhui Ma, Hui Chen, Shu Zhan, Jing Cao. Boron Removal Method, Technology and Process for Producing Solar Grade Silicon by Metallurgical Method. Progress in Chemistry, 2022, 34(4): 926-949 DOI:10.7536/PC210710

登录浏览全文

4963

注册一个新账户 忘记密码

1 Introduction

Energy is an important material basis for the development of national economy and the improvement of people's living standards. However, the energy crisis and environmental degradation that we are facing today seriously restrict the sustainable development of national economy and society. Therefore, the development and utilization of new energy is imminent. Solar energy, as an important and effective renewable energy, has attracted wide attention because of its inexhaustible, inexhaustible and no environmental pollution. Vigorously developing solar photovoltaic power generation is an effective way to solve the future energy shortage. Silicon-based solar cells accounted for more than 95% of the photovoltaic market in 2019 due to their low cost, higher photoelectric conversion efficiency and better material stability[1]. In order to promote the utilization of solar energy and meet the growing demand for silicon-based solar cells, it is essential to produce high-quality silicon.
According to purity and industrial application, silicon materials can be divided into 3 grades: metallurgical grade silicon (MG-Si, purity: 99% (2 N)), solar grade silicon (SOG-Si, purity: 99.9999% (6 N)) and electronic grade silicon (EG-Si, purity 99.9999999% (9 N)). SOG-Si and EG-Si can be used to produce solar cells, while MG-Si has been the raw material for the production of SOG-Si and EG-Si. Waste generated from EG-Si was previously the main supply route for SOG-Si production. Due to the rapid development of the photovoltaic market in the past few decades, there is a shortage of waste silicon powder, which is now mainly from non-high-quality polysilicon. High purity silicon is produced by two main techniques: the Siemens method and the metallurgical method.
Metallurgical preparation of solar grade polysilicon is a production method that effectively reduces production costs and is specifically targeted at solar grade polysilicon, which is conducive to the large-scale production of independent polysilicon in China, reduces the cost of photovoltaic power generation, and promotes the rapid development and overall improvement of the level of the solar cell industry. Various metallurgical purification processes in metallurgy include slag treatment, wet leaching, vacuum evaporation, directional solidification, electron beam, plasma, solvent refining, etc.
With the development of solar cells, the requirement of B removal in the production process is getting higher and higher. In fact, when the concentration of B is higher than 0.3 ppmw, the lifetime of minority carriers will be sharply shortened due to B-O defects, which will seriously affect the photoelectric conversion efficiency of the cell, so the content of B in the cell must be strictly controlled[2]. However, due to the large segregation coefficient of B and the lower saturated vapor pressure than Si, B can not be effectively removed only by directional solidification and vacuum melting. At present, the main processes of boron removal are slag refining, low temperature silicon-based solvent refining, gas blowing, high temperature plasma refining and acid leaching. In order to effectively reduce B concentration in large-scale SOG-Si production process, it is important to review the development of refining B removal process.

2 Properties of boron in silicon

2.1 Solubility of boron in silicon

Tang et al. Developed a thermochemical database covering 33 elements[3]. The Si-B binary system is shown in Fig. 1, and the composition of B in SOG-Si generally ranges from ppm to a few percent[4⇓⇓⇓~8]. However, given their validation in binary alloy systems, caution is warranted in the use of these data.

2.2 Diffusion coefficient of boron in silicon

Refined Si for solar cell applications, the dissolved B will be at the ppb ~ ppm level. Therefore, the mass transfer of B from the bulk phase of Si to the reaction plane/interface is an important step in the refining process, and the diffusion coefficient, especially the diffusion coefficient measured by experiments, is very necessary for the discussion of kinetics and the realization of modeling. Fig. 2 shows the diffusion coefficient of B in solid and liquid Si as a function of temperature[9⇓⇓⇓⇓⇓⇓⇓⇓⇓~19]. Here, we recommend the results of Tang et al. And give the relationship between the diffusion coefficient of B and temperature, as shown in Equation (1)[20].
l g D = - 10373 T - 8.81 ( 1000 K < T < 1678 K ) - 7.77 ( ± 0.15 ) ( 1687 K < T )

2.3 Diffusion coefficient and mass transfer coefficient of boron in silicon and slag

Slag treatment can remove B from MG-Si with high removal efficiency. However, single-stage slag treatment does not produce silicon with a sufficiently low B concentration, and the removal rate depends on the reactor/process design. Therefore, it is interesting to identify the limiting step for B removal from MG-Si by slag treatment. Both the mass transfer of boron (B) in molten Si and the mass transfer of boron oxide (BO1.5) in slag (Table 1) affect the removal rate of B during slag refining.

2.4 Activity coefficient of boron in silicon and slag

In the field of boron removal, it is of great significance to clarify the mechanism and kinetics of boron removal process for improving the efficiency of boron removal. Activity is also one of the thermodynamic properties to achieve numerical simulation and model building.
The activity coefficient of B is expressed by the first-order interaction coefficient, as shown in equation (2).
l g γ B = l g γ 0 B + ∑ 1 i = n ε i B x i
Where γ 0 B is the interaction coefficient of xi and ε i B 分 别 表 示 i 物 种 的 摩 尔 分 数 和 i species with B in molten Si relative to the activity coefficient of B in molten Si in pure liquid infinite dilute solution.
Noguchi et al. Reported the activity coefficient in Si melt determined by equilibrating solid BN and liquid Si under N2 atmosphere from 1723 to 1923 K, as shown in Equation (3)[21].
l g γ 0 B = - 11100 T + 5.82 ( 1723 K < T < 1923 K )
Tanahashi et al. Reported the activity coefficient in Si melt determined by equilibrating solid BN and liquid Si under N2 atmosphere from 1723 to 1923 K, as shown in Equation (4)[31].
l g γ 0 B = 2.5 ± 0.2 ( 1723 K ) 2.3 ± 0.2 ( 1773 K )
Inoue et al. Studied the effect of Ca on the thermodynamic properties of B in Si melt, and the activity coefficient obtained is shown in formula (5)[32].
l n γ 0 B = 2.27 ( ± 0.01 ) ( 1723 K )
Yoshikawa et al. Determined the activity coefficient by equilibrating molten Si with solid Si3N4 and BN at 1693 K and 1923 K[33]. The activity coefficient of B in solid Si is measured by the segregation behavior of B between solid Si and Si-Al melt, as shown in equations (6) and (7)[34].
l g γ 0 B = 289 ( ± 450 ) T + 1.19 ( ± 0.25 ) ( 1693 K < T < 1923 K )
l g γ 0 B = 14073 ( ± 276.6 ) T - 5.73 ( ± 0.19 ) ( 1273 K < T < 1473 K )
Khajavi et al. Used the activity coefficient of B in Si-Fe alloy to evaluate the activity coefficient in solid Si, as shown in equation (8)[35].
l g γ 0 B = 16317 ( ± 282 ) T - 7.06 ( ± 0.18 ) ( 1483 K < T < 1583 K )
The above results are shown in Fig. 3. One possible reason why the data of Noguchi et al. Are different from those of other researchers is that they did not consider the formation of Si3N4. When these data are ignored, the activity coefficient of B can be expressed by equation (9)[21].
l g γ 0 B = 17450 T - 8.15 ( 1173 K < T < 1923 K )
Table 2 lists the interaction coefficients of different species in molten Si with B, and the self-interaction coefficient of B in Si can be fitted by equation (10).
ε B B = 2153860 T - 1435.47 ( 1723 K < T < 1773 K )
Teixeira et al. Studied the basicity (the mass fraction ratio of CaO-SiO2 to SiO2, denoted as R) and activity coefficient of BO1.5 of CaO-SiO2 and CaO-SiO2-CaF2 slag systems at 1823 K, as shown in Equation (11)[36].
l n γ B O 1.5 = - 4.00 R + 3.67 ( 1823 K )

2.5 Segregation coefficient of boron between solid silicon and solvent

During solidification, a well-defined interface between solid Si and liquid melt can be observed. At the solid/liquid interface, the B concentration in the solidified Si is in equilibrium with that in the solvent. Taking the solidification of Al-45 mol% Si melt as an example, the first phase precipitated is solid Si, which is called primary Si. The solubility of B in liquid is higher than that in solid Si, so it can be effectively removed by solvent refining. Of course, a certain amount of Al remains in the primary Si and must be removed in a subsequent stage.
The segregation coefficient (kB) of B can be expressed by equation (12).
k B = x B S x B L = γ B L γ B S e x p Δ G 0 B R T
Where x B S and x B L are the mole fractions of B in the solid and liquid phases, respectively, γ B S and γ B L are the corresponding activity coefficients, and Δ G 0 B refers to the Gibbs energy of B melting. Equation (13) represents the kB between solid Si and Si-Al melt measured by the temperature gradient zone melting method[33]. Table 3 lists the kB values for several solvents.
l n k B = 4.23 - 7300 / T

2.6 Technical difficulties and challenges of boron removal from silicon

Boron and oxygen will form boron-oxygen pairs, which will reduce the lifetime of minority carriers and damage the photoelectric conversion efficiency of solar cells. Therefore, the B impurity content in solar grade silicon is generally required to be less than 0.3 ppmw. However, because the segregation coefficient of B in Si is 0. 8 (1687 K), which is close to 1, B can not be separated from Si by directional solidification. In addition, the melting point of B is about 2573 K, which is higher than that of Si, and the saturated vapor pressure at high temperature (only 10-6Pa at 1773 K) is far lower than that of Si, so it can not be removed by vacuum melting. Although the removal of B by pickling has a certain effect, it is difficult to reduce to less than 1 ppm. The removal of B by plasma melting has a better effect, but the cost is high and the industrialization is difficult. In the process of slagging refining, the content of B in Si in the equilibrium state will increase with the decrease of temperature, because in the cooling process, the back diffusion of B in slag occurs, and some B impurities enter Si, resulting in the increase of B content in Si. In addition, the removal effect of B is related to the oxygen potential of the slag agent and the basicity, and the oxygen potential plays an important role in the oxidation removal of B in the case of low basicity; When the basicity is too large, the effect of basicity on B is stronger than that of oxygen potential, and the oxidation effect of slag on B in the alloy is weakened.
Impurities in Si are thermodynamically unstable at low temperature, that is, the solid solubility of impurities decreases with the decrease of temperature. For non-metallic impurities B and P, high energy consumption methods such as multiple slagging refining and electron beam smelting are generally used to remove them in industry, but there is still no alternative method with low energy consumption, which also causes problems such as high energy consumption, long production cycle and unstable product quality in metallurgical process.

3 Boron removal process

There are five main processes for B removal: gas blowing, slag treatment, plasma, acid leaching and solvent refining.

3.1 Boron removal by gas blowing

As shown in Fig. 4, blowing can be defined as blowing an oxidizing gas (such as H2O, O2 and H2) into the liquid Si (or slag phase when combined with slag treatment) at high temperature,And B contained in Si is oxidized to an oxide such as BO, B2O3, B2O,BO2 and B2O2) or under a humid atmosphere (e.g. Ar-H2O) forming hydroxide gases (e.g. BHO,Processes of BHO2, BH2O2, and U BH3O3. These gaseous products are easily vaporized because their equilibrium partial pressure is 1 times UNK of the oxidizing gases mentioned above.
In this paper, the mechanism of B removal by steam and oxygen refining is described in detail, as shown in Fig. 5. The water vapor and oxygen are usually mixed with the carrier gas Ar before blowing. B is mainly oxidized into gaseous boride HBO, which can escape into the gas phase at high temperature of the melt, and the metal impurities will also be transformed into the corresponding oxides.

3.1.1 Blowing with wet hydrogen

Mixed wet hydrogen (Ar-H2O-H2 mixed gas) was used as the reaction gas to remove B from the Si liquid. B in the Si melt is extracted into the gas phase through the vapor-liquid interfacial reaction. In addition to B, Si can also be oxidized by water vapor, and the reaction is represented by formulas (14 to 16):[43,44]
B + H 2 O ( g ) = H B O ( g ) + 1 / 2 H 2 ( g )
S i ( l ) + H 2 O ( g ) = S i O ( g ) + H 2 ( g )
B + S i O ( g ) + 1 / 2 H 2 ( g ) = H B O ( g ) + S i ( l )
HBO and SiO are the main boron- and oxygen-containing gases in contact with the silicon melt. MG-Si melt was refined by the mixed gas of H2O and H2, and the results showed that B in silicon combined with H2 and intermediate SiO to form gaseous HBO. In addition to the formation of gas-phase HBO, impurities are also oxidized and enter the slag phase. Thermodynamic results show that gaseous volatiles are not the dominant form of B division[43,45]. Tangstad et al. Used FactSage 5.0 software to calculate the gas partial pressure of H2-3.2%H2O-Si-B system with B/Si ratio of 10-5(10 ppmw) at different temperatures[44,46]. In terms of kinetics, the overall mass transfer coefficient was (51.1 ± 1.8) μm/s when water vapor and hydrogen were blown onto the surface of the melt in a 70 mm diameter crucible at a flow rate of 16 L/min. It is confirmed that the removal rate of B increases with the increase of gas flow rate and crucible diameter, and the bulk gas mass transfer is found to be the only rate-determining step. Therefore, the removal of B by oxidizing gases such as steam in hydrogen at high temperature may be very effective and should be further studied[44].
The study on the limiting step of boron removal in Si purification based on the gas blowing method shows that the chemical reaction on the surface of Si melt is close to chemical equilibrium. The mass transfer rate in the gas phase near the interface is identified as the rate-limiting step[47]. Meanwhile, hydrogen and/or water vapor is used as a reaction gas for the plasma blowing method, and B is volatilized by simultaneously adding oxygen and hydrogen[48,49]. Under the condition of ultra-high temperature plasma method with electromagnetic casting (EMC) process, B volatilizes in the form of BH and BH2 when hydrogen alone is used as the reaction gas. Water vapor, when used alone, decomposes into hydrogen and oxygen. Hydrogen and oxygen atoms react with B in Si melt to form volatile borides BOH, B(OH)x, etc. According to the temperature dependence of the standard Gibbs free energy in the B-H2O system, BOH is 10 times more volatile than BO and is the most stable above 2000 K. Therefore, not only hydrogen but also oxygen blowing is required to effectively remove B from Si melt.

3.1.2 Blowing with wet oxygen

The addition of water vapor to oxygen results in the formation of boron hydrates with a more negative standard Gibbs free energy than boron oxides. At the same time, the volatility of boron hydrate is higher than that of boron oxide. Wu et al. Studied the thermodynamics of B removal from MG-Si by pure oxygen and wet oxygen blowing, and the representative reaction is given by equation (17):[50]
4 B + 2 H 2 O ( g ) + O 2 = 4 H B O ( g )
This implies a relationship between the amount of B divided and the partial pressures of H2O and O2. The results show that the partial pressure of all gaseous boron hydrates in the B-H2O-O2 system is nearly 105 times higher than that of gaseous boron oxides in the B-O2 system[50]. Therefore, the effect of removing B by wet oxygen blowing is better. When the concentration of B in silicon is greater than 4.2 ppmw, B is mainly volatilized in the form of gaseous B3H3O6. At lower concentrations, gaseous HBO2 becomes an important volatile species[50]. These thermodynamic results are consistent with the results of Nordstrand et al. And Lee et al[46][48]. When the B concentration in refined Si is 0.1 ppmw at 1723 K, the equilibrium partial pressure of HBO2 will reach 2500 Pa, which indicates that this boron hydrate has good volatility. At higher temperatures, the volatility of gaseous boron hydrate decreases slightly, while the opposite is true for gaseous boron oxide. In order to verify the thermodynamic results and determine the optimal H2O-O2 composition ratio, dry oxygen was first blown into the molten MG-Si in the ladle and purified at 1773 K for 2 H. The experiment was conducted at an 800 kg scale using compressed oxygen blown into a ladle lined with magnesia-alumina refractory bricks at a rate of 8 L/min. A 120 G sample of the refined silicon was then removed and placed in a water-cooled copper crucible in a DC arc furnace. The results showed that the B concentration in Si was reduced from 18 ppmw to 2 ppmw (89% efficiency) at a 1:1 H2O:O2 gas composition[50]. When the gas composition is 2:1, the B concentration is further reduced to 0.6 ppmw, and the B removal efficiency reaches 96.7%[51].

3.1.3 Blowing with other moist gases

Similarly, carbon dioxide is used and tries to remove B. Suzuki et al. Used Ar-0.44%H2O-0.44%CO2 mixed gas to refine MG-Si[52]. Meanwhile, Ar-1.24%H2O gas was provided as a control. The reaction is shown in formula (18).
2 B + H 2 O + C O 2 = 2 H B O ( g ) + C O ( g )
The results showed that the addition of 0.44%CO2 to the Ar-0.44%H2O gas did not improve the removal of B. At two different melting times, the B removal rate of Ar-1.24%H2O is 10% – 15% higher than that of Ar-0.44%H2O-0.44%CO2, which indicates that a SiO2 film is formed on the surface of Si melt. When H2O is present in the working gas, it is dissociated into H and O atoms, while CO2 decomposes into CO and O atoms. These unstable H or CO atoms may react with the possibly formed SiO2 to form volatile SiO before recombination, thus cleaning the surface of molten Si[52].
In order to predict the thermodynamic trend of gaseous borides, the mixing free energy of representative borides BH3, BO and HBO in the Ar-H2O-(H2,O2 or CO2)-B system was calculated by thermodynamic software (HSC Chemistry), as shown in Fig. 6. It is clear that single H2 or CO2 gas blowing refining has little effect on B removal, while both oxygen and steam are effective for B removal. The most effective component is water vapor added to oxygen, and the gaseous boron product HBO is most readily formed when the chemical composition is a mixture of 2H2O-O2.
At present, there are still the following problems in the industrial application: First, the removal effect of B is still difficult to meet the requirements of solar grade polysilicon. Secondly, there are some limitations in the way of introducing gas, and the ventilation parts are easy to corrode or damage. Considering the viscosity of Si liquid, the diffusion effect and reaction degree of gas are not ideal. In addition, it is inevitable that a part of Si will be oxidized into silicon dioxide and silicon monoxide due to its strong reducibility, resulting in a small amount of Si loss. In addition, the saturated vapor pressure of borides is normally very low, which limits the reaction rate of B removal.

3.2 Boron removal by slag treatment

Slag treatment process refers to adding refining agent with melting point higher than Si into metallurgical grade silicon to oxidize the impurity elements in liquid Si, so that the product enters the slag phase, and the metal and slag reach thermal equilibrium, thus achieving the purpose of removing impurities. The multi-component slag composed of oxides can absorb or accumulate impurities during refining, and can prevent the contamination of Si and heat loss. B is oxidized by the slag at the phase boundary, and after sufficient refining time, the interfacial reaction reaches equilibrium. Si with low impurity content can be further separated and purified from the refining slag. As shown in Fig. 7, the B removal process during CaO-SiO2 slag treatment can be divided into five steps: (1) mass transfer from the Si bulk phase to the metal boundary layer, (2) diffusion through the metal boundary layer to the interface, (3) interfacial reaction occurring at the interface between the metal and the slag, (4) diffusion from the interface through the slag boundary layer, and (5) mass transfer from the slag boundary layer to the slag bulk phase[53].
The interfacial reaction can be expressed by equation (19). Its equilibrium constant K is given by equation (20).
B + 3 / 4 S i O 2 = B O 1.5 + 3 / 4 S i
K = γ ( B O 1.5 ) x ( B O 1.5 ) α [ S i ] 3 / 4 γ [ B ] x [ B ] α [ S i O 2 ] 3 / 4
Where xi, αi, and γi are the mole fraction, activity, and activity coefficient of substance I, respectively.
The distribution coefficient (LB) is usually used to express the removal efficiency, as shown in Equation (21).
L B = x ( B O 1.5 ) x [ B ] = K · γ [ B ] · α ( S i O 2 ) 3 / 4 γ ( B O 1.5 ) · α [ S i ] 3 / 4
It should be noted that this value can only be used when the reaction has reached equilibrium, and it depends only on the slag composition, temperature and atmosphere. The LB should be as high as possible to minimize the amount of slag required for refining, as secondary or multiple slag treatments are unfavorable for industrial applications. The typical value of LB is between 1 and 2, while the highest value of LB reported by Teixeira et al. Is only about 5.5, which means that refining Si requires a large slag-to-silicon ratio (μ), but excessive slag consumption is very unfavorable to the refining process[36]. The constraint of the maximum LB value has prompted scholars to select ternary or multi-component slags with suitable thermodynamic and kinetic properties (such as low viscosity, low melting point and high basicity).
The experimental results of refining with CaO-SiO2 slag show that the boron distribution ratio and boron removal efficiency increase significantly with the increase of CaO content in slag. The distribution ratio of B varies parabolically with the increase of the mass ratio of CaO/SiO2. When the slag composition is 60%CaO-40%SiO2, the maximum B distribution ratio can reach 1.57. When the slag/MG-Si molar ratio is 2.5, the B concentration in the refined Si is reduced from 18 ppmw to 1.8 ppmw at 1823 K for 3 H, and the B removal rate is 90%. Wu et al. Studied the effect of basicity (Λ) of CaO-SiO2 slag on the distribution ratio of B, and defined and calculated the basicity Λ of molten slag by equation (22)[54].
Λ s l a g = ∑ x i n i Λ i ∑ x i n i
Where xi is the mole fraction of oxide component I, ni is the number of oxygen atoms in each oxide component I, and Λi is the theoretical basicity of component I. As shown in Fig. 8, in the phase diagram of the CaO-SiO2 system, since the CaO-SiO2 slag in this composition region has a lower liquidus, the selection of the low CaO region is more suitable for purifying silicon. The slag basicity considered in this review ranges from 0.5 to 4.5. It is concluded that the rate-determining step of boron removal is the mass transfer of boron from molten Si to the slag/silicon interface, or the mass transfer of B2O3 from the slag/silicon interface to the slag[55].
In addition to the CaO-SiO2 binary slag,Researchers have also used ternary or multi-component slag systems such as CaO-MgO-SiO2, Al2O3-CaO-SiO2, Al2O3-CaO-MgO-SiO2, K2CO3-CaO-SiO2 and CaO-SiO2-Na2O-Al2O3 to remove B from MG-Si.And some more fruitful results have been achieved[56][57,58][59,60][61][62].
The metal-slag mass transfer coefficient ranges from 1.7 μm/s (for the slag composed of the same amount of 50%SiO2-50%CaO) to 4.3 μm/s (for the slag composed of 40%SiO2-40%CaO-20%MgO), and it is also found that the slag with low melting point and high basicity has better B removal effect[23]. The researchers also investigated the removal of B from MG-Si using Na2O-CaO-SiO2 slag[63,64]. The slag treatment process was carried out in two passes, and directional solidification was performed to promote the separation of slag from silicon. It is considered that the higher the refining temperature is, the higher the Na2O concentration in slag is, and the higher B removal efficiency is[65]. It has been proposed that the removal of B from the Si liquid phase is achieved by the oxidation of B with Na2O at the interface of slag and Si, and the oxidized B is further gasified from the slag by the formation of sodium metaborate Na2B2O4 at the interface of slag and gas[22]. The Na2CO3-Al2O3-SiO2 sodium base refining slag system was used as the raw material for boron removal from silicon bath. The results show that the addition of a proper amount of Al2O3 into the slag is beneficial to maintaining the basicity of the slag, thus improving the boron removal efficiency[66]. White et al. Studied the mass transfer rate of B between liquid Si and CaO-SiO2 slag under stirring at 1823 K, and the removal rate of B in Si was improved by mechanical mixing[67]. By emulsifying the slag phase, the reduction of calcium oxide at the interface seems to greatly enhance the mechanical mixing effect.
The distribution ratio of boron is affected by the basicity and oxygen potential of the slag, and the removal efficiency of boron in the slag treatment depends on the boron content in the slag. There is always a significant difference in viscosity and melting point between the slag and Si. During the slag refining process, it is difficult for slag particles to fully contact with liquid Si. Therefore, it is of great significance to improve the B removal ability of slag by selecting the appropriate basicity and oxygen potential of slag to accelerate the mass transfer of B between Si liquid and slag. Both the calculated and experimental results show that the basicity of the slag has a great influence on the distribution ratio of B[27]. In another study, ZnO was added to the calcium silicate slag to improve the B removal efficiency due to its stronger B removal ability compared with SiO2[68,69]. The mechanism of removing B from MG-Si by CaO-SiO2-ZnO ternary slag system refining is shown in Fig. 9. The process of removing B can be described as follows: (1) B dissolved in Si solution diffuses to the slag/silicon interface and/or slag components diffuse to the interface; (2) chemical reaction between B and slag at the interface; (3) Borate diffusion at the interface and Zn volatilization. When 8% ZnO was added into 50%CaO-50%SiO2 calcium silicate slag, the B concentration in refined Si was reduced from 12. 9 ppmw to 2. 2 ppmw at 1723 K, the distribution ratio was 3. 04, and the B removal rate was 83%[69]. After distillation treatment, the residual Zn in Si is less than 0.05 ppmw.
It is found that the composition of the slag system has an important effect on the B distribution ratio. The addition of Na2O with basicity higher than that of other oxides can significantly improve the overall B capacity of slag[57,62,63]. This result has been confirmed by almost all studies: the higher the basicity of the slag system, the more favorable it is to remove B. The heating mode is the secondary factor affecting the effect of B removal, and also affects the distribution ratio of B. Because induction heating has the function of electromagnetic stirring, it is considered that induction heating is more beneficial to B removal than resistance heating. Generally, higher refining temperature will enhance the kinetic mass transfer and reduce the viscosity. However, in slag treatment, increasing temperature has no significant effect on the distribution ratio of B.
It can not be ignored that in the process of slag treatment, there is a certain amount of Si inclusions in the slag, which will cause the loss of some Si. At the same time, the volatilization of SiO produced by the reaction between SiO2 and Si in the slag can also cause the loss of Si. In addition, due to the strong affinity of basic oxides such as CaO and MgO with SiO2, too much of these oxides will reduce the activity of SiO2 while increasing the basicity of slag, resulting in the reduction of oxygen potential of slag, which is not conducive to the removal of impurities.
The slag treatment process adds new impurities in the slagging process, which need to be further removed in the subsequent process, thus increasing the process steps. A large number of experiments and comparisons are needed to select a suitable slag former, which can not only effectively react with the impurities in the Si melt to form a stable slag phase, but also do not bring in impurities that are not easy to remove in the subsequent refining, and are easy to remove after the directional solidification of the Si melt. And the effect of impurity removal in slagging is closely related to reaction kinetics, impurity diffusion and impurity separation coefficient. The selection of slagging agent is the key to the purification of Si by slagging refining. In addition, the use of large amount of slag is conducive to obtaining good removal effect, but how to deal with a large amount of tailings is also a very important issue.
Although slag treatment has obvious advantages such as low cost and large scale, considering the limitation of B removal, the process can be comprehensively utilized to improve the kinetic conditions in combination with gas blowing.

3.3 Boron removal by combined refining of gas blowing and slag treatment

In the refining process of MG-Si, the combined refining of gas blowing and slag treatment can be used to remove B from silicon in ladle. B in Si as well as oxides or borates are removed in the form of gaseous borides. The SiO2 produced by oxygen blowing can promote the oxidation of boron in Si melt, and the adsorption of boron oxide by slag can accelerate the volatilization of gaseous boride. This synergistic relationship has been proposed[70,71]. A mixture of 0.5 – 2 MPa of Ar, O2 and water vapor was blown into the MG-Si melt to remove B, and the technology of gas blowing in the ladle combined with slag treatment is shown in Fig. 10. It is clear that a high distribution ratio of B between slag and Si must be achieved to maximize the oxygen potential (a function of silica activity) and slag basicity (a function of slag oxygen ion O2- activity). Fig. 11 shows the partitioning of B between slag and Si versus equilibrium slag composition. Liquid boron oxides/borates and gaseous borides are the most important forms of boron removal from MG-Si by the combined gas blowing and slag treatment technique[72].
The H2O-O2 gas mixture and chlorinated CaO-SiO2-CaCl2 slag were combined to separate B from molten MG-Si, and the synergistic separation behavior of B is shown in Fig. 12. The liquid species B2O3 and the gaseous species BO, HBO, BClO can be produced simultaneously by the combined use of Ar-O2-H2O gas blowing and CaO-SiO2-CaCl2 slag treatment processes. Boron oxide, the intermediate product, is immediately adsorbed by the alkaline calcium silicate slag to form a new boron-calcium slag phase (3CaO·B2O3), which rapidly reduces the B concentration at the slag-silicon interface and enhances the driving force for B diffusion from the Si melt to the slag/silicon interface or gas/silicon interface. Therefore, it is considered that the formation of calcium borate can accelerate the formation and volatilization of gaseous products BO, HBO and BClO.
The B in MG-Si was removed at different refining times by using 40%Ar-60%O2,40%Ar-60%H2O and 40%Ar-20%O2-40%H2O gas mixture to prepare 42.5%CaO-42.5%SiO2-15%CaCl2 slag at 1823 K. The results show that the efficiency of B removal is greatly improved by the combination of gas blowing and slag treatment. When the anhydrous mixture of 40%Ar-60%O2 reacted with 42.5%CaO-42.5%SiO2-15%CaCl2 slag at 1823 K for 3 H, the B concentration in Si decreased from 22 ppmw to about 2 ppmw (B removal efficiency was 91%). However, when 40%Ar-20%O2-40%H2O gas blowing was used, the B level was reduced to 0.75 ppmw and the B removal efficiency reached 96.6%.
The research on the combined refining method of gas blowing and slag treatment has a long history, and the research results of scholars at home and abroad are constantly improving, and the removal effect of impurity B is more and more close to the requirements of solar grade silicon materials. However, when the slag and Si are mixed and stirred and an oxidizing gas is blown in, since Si is partially oxidized to SiO2 and absorbed by the slag, the viscosity of the slag increases, which reduces the mixing efficiency between the slag, Si, and the gas, and the oxidation reaction rate of B in Si is slowed down. Moreover, part of the oxidizing gas will react with Si at high temperature. In order to reduce the loss of Si, the experimental temperature should be controlled at a specific temperature range, which requires a higher process of the equipment. The blowing process will inevitably cause the oxidation and volatilization of Si, thus reducing the direct recovery rate of Si, which is a problem that can not be ignored in the application of blowing oxidation refining in the production of solar grade silicon by metallurgical process, and should be paid attention to by the industry.

3.4 Boron removal by plasma refining

Similar to gas blowing, plasma refining is an efficient metallurgical process that uses the high reactivity of the reaction gas to remove impurities[73,74]. Generally, the reaction gas is composed of O2, H2, and H2O molecules, which are decomposed into atoms or ions O, H, OH, O+, OH+ by the plasma gas at high temperature. By plasma refining with O2 or H2 or H2O reactive gases, the B impurities in MG-Si will be transformed into gaseous borides, such as BO, B2O, HBO or BH2, etc. B atoms in molten Si are removed through three processes: (1) B in molten Si migrates to the liquid surface; (2) production of volatile boride by chemical reaction on the gas-liquid surface; (3) The plasma torch migrates the boride away from the gas-liquid surface. This way of transport through the plasma gas is also important for bringing the reaction gas to the liquid surface. The mass transfer rate in Si liquid is considered to be the rate-determining step in kinetics, because the rate of chemical reaction is usually much faster than the mass transfer rate at high temperature.
The plasma containing 4.3%~4.6%H2 and 50.4%~50.6%H2O was blown into the MG-Si melt with B concentration of 5 – 10 ppmw, and the mechanism of the deionization process is shown in Fig. 13[75]. Similar to the gas blowing process, the chemical reactions of plasma refining are shown in equations (23) and (24):
B + H 2 O ( g ) = B O ( g ) + H 2 ( g )
B + H 2 O ( g ) = H B O ( g ) + 1 / 2 H 2 ( g )
The results show that the content of B in refined Si is reduced to less than 0. 1 ppmw after refining for 400 min[75]. In addition, the researchers found that the removal rate of B is proportional to the water vapor content in the plasma gas according to the reaction kinetics. The increase of water vapor in the plasma gas not only accelerates the reaction rate, but also improves the removal efficiency of B[76].
Yuge et al. Combined vacuum, electron beam, plasma, gas blowing, directional solidification refining and other technologies to form a comprehensive metallurgical process for solar grade polysilicon[77]. In the vacuum system, the MG-Si was heated to the molten state by electron beam to remove P impurities, and then further processed by directional solidification. MG-Si was heated with a plasma torch in an Ar atmosphere, refined by a reactive gas mixture of O2 and H2O to remove most of the B and C impurities, and then subjected to a secondary directional solidification treatment. The concentration of P, B and metal impurities in the finished product is less than 0.1 ppmw, which meets the requirements of solar grade polysilicon.
Plasma refining has the advantages of fast material reaction, high temperature and high electrothermal conversion efficiency, and has a good removal effect on B. However, due to the concentration of heat provided by the plasma arc, it is easy to heat the charge unevenly. At present, the experiment of plasma refining method has made good progress, but from the perspective of industrial production, the operation of plasma refining method is relatively complex, and the reaction progress can not be controlled, and the oxidation of plasma weakens the volatilization of impurity B, and the high temperature environment of plasma will promote the volatilization loss of main Si. In addition, the technology has the inherent disadvantages of expensive refining equipment, high cost, high energy consumption, low stability and difficult maintenance, which determines the limitation of large-scale popularization and application of plasma refining method in industrial production.
In view of the above problems, plasma diagnosis and numerical simulation can be combined to measure and calculate the axial velocity distribution, temperature distribution and current density distribution on the surface of the molten pool in the smelting process, so as to find the optimal arc current and atmosphere pressure range, improve the smelting efficiency and enhance the product quality. In addition, it is necessary to optimize the design of plasma equipment to improve the life of components and arc stability during production, reduce energy consumption while improving production continuity, and make it better serve the actual production.

3.5 Boron removal by acid leaching

Metal impurities that can form intermetallic compound phases (such as silicides and silicates) with Si are always segregated at grain boundaries, and the extraction of Si by acid leaching has a long history. The ElKem Silgrain® process, which is still commercially producing MG-Si, immerses powders of 90FeSi alloy (90% Si) in an acidic solution of HCl-FeCl3, in which impurities at grain boundaries continuously react with the acidic mixture and then decompose[78]. This process is particularly effective in removing transition metals (e.g., Fe, Cu) because of the low segregation coefficient of these impurities in Si.
It is well known that surface treatment (or leaching) with HF or HCl/HF mixtures can also wash out residual impurities in the form of oxides or silicates. Grinding Si prior to acid leaching to increase surface exposure is a method of enhanced purification[79,80].
Sun et al. Investigated the effects of hydrofluoric acid concentration and Si particle size on B removal[81]. The boron content in Si powder with particle size of 44 ~ 74 μm can be reduced from 128 ppmw to 10. 9 ppmw by calcination at 1273 K for 2 H, vacuum quenching at 373 K and acid leaching at 2 h,323 K HNO3+HF for 80 min. In addition, the B concentration in Si powder (particle size 74 – 154 ppmw) was reduced from 8.6 ppmw to 3.0 ppmw by HF+H2O2+(CH2OH)2 acid leaching at 328 K for 2 H[82].
It must be emphasized, however, that this method does not actually refine Si at B concentrations below its solubility limit at the corresponding solidification composition. But fortunately, in solidified MG-Si, B appears at the grain boundaries in the form of additional phases, which can be removed by crushing and subsequent acid leaching, resulting in Si particles of about 2 mm[83].
Although the acid leaching process produces a large amount of solvent, it is not a very energy-intensive or costly method. In fact, solvent metals and acids can be converted into valuable by-products such as AlCl3 and FeCl3. These by-products can be sold at a reasonable price to recover most of the cost of the acid leaching process (i.e., the cost of solvent metal and acid). Therefore, the cost of the acid leaching process used in industrial production is actually quite low.
The acid leaching process has the advantages of simple equipment, low energy consumption and large-scale operation, and has good removal effect on most metal impurities, while for the non-metallic impurities with large segregation coefficient, B and P exist in the form of replacing Si atoms or in the interstitial position of the lattice, and the removal effect of wet leaching on them is not obvious. In addition, it is found that when hydrofluoric acid is used for pickling, a large amount of Si powder will float up and be difficult to sink, which makes it difficult to stir and cause a large loss of Si powder. Too fine particle size not only increases energy consumption, but also brings inconvenience to subsequent processes.
There is a limit to remove impurities from Si by pickling, because the distribution of impurities in Si is not uniform and the removal is incomplete. In addition, even at the time of pulverization, impurities may be completely coated with Si and do not come into contact with the outside world, making it difficult for them to react with acid. This method has long treatment time, limited purification effect and low efficiency, so pickling can only be used as a pretreatment link in the purification process of metallurgical method to prepare for further purification.

3.6 Boron removal by solvent refining

Solvent refining is carried out on the basis of fractional crystallization. At about 1473 K, Si is mixed with low melting point metals (such as Al, Sn, Cu, Zn) to form an alloy melt. During directional solidification, the impurities of B accumulate in the molten eutectic phase (such as Si-Al, Si-Sn, Si-Cu and Si-Al-Sn) because the liquid melt has higher solubility than solid Si[84][85][86][87]. During directional solidification with a slightly convex solidification front (instead of a planar front), a clear interface between solid Si and liquid melt can be observed, as shown in Fig. 14. By definition, ideal solvents are those with low melting point and/or high solubility of B.
The metal used to purify Si should meet the following criteria[88].
(1) No intermediate compound is formed. Some silicon-based alloy systems produce intermediate compounds such as Si-Ti silicides, which are difficult to remove by acid leaching and lead to serious entrainment of metal impurities.
(2) Small values of ki. In the cooling stage, Si is recrystallized and impurities are redistributed between the solid and liquid phases according to the ki, so a small ki value can ensure a high purification efficiency.
(3) The Si content in the eutectic point is low. The solidified alloy contains primary Si and eutectic Si metal. Si is difficult to separate from the metal, and its homogeneous microstructure can lead to severe Si loss. Therefore, a lower Si concentration at the eutectic point can reduce the depletion of Si.
(4) Primary Si is easily separated from the alloy. The significant difference in properties between Si and metals makes it easy to isolate Si.
(5) Economical and common metals. In order to reduce the cost, the metal used should be economical and widely used, because the demand for this metal is considerable.
According to the above standards and the physical properties of common metals, several metals such as Al, Sn, Zn, Cu, Fe and In meet the requirements. The following sections discuss the processes involved. In addition, the effects of solvent refining process parameters, such as cooling rate, heating mode and treatment time, on the segregation behavior of B and the purification efficiency of Si were discussed.

3.6.1 Si-Al alloy

Thermodynamic studies on the solidification of Si from solvents have been carried out mainly on Si-Al solvents[89⇓~91]. The basic steps of solvent refining are: (1) premelting one or more low melting point metals and forming a Si-rich solvent or alloy with MG-Si; (2) Less contaminated Si crystals can be crystallized from the mixed solvent or alloy by cooling the solvent or pulling out the solvent from the heating zone, and the impurities are driven and enriched into the solvent or alloy. The process can continue until the temperature reaches the eutectic point of the Si alloy; (3) separating the purified Si crystal from the solvent or alloy. Fig. 15 is a schematic diagram showing the interaction of B with additive elements in a Si-Al melt[92]. In general, the additive elements have a strong affinity for B in Si-Al melt and will form borides. B is trapped in these compounds so that its concentration in the remaining melt is reduced. In some cases, the primary silicon phase grows during solidification. These compounds become part of the primary silicon phase, in which the B content increases (Figure 15B). In other cases, the B content in primary silicon decreases if these compounds are not absorbed (Figure 15 C). If the additive element has no affinity for B in Si-Al melt, the additive element does not affect the concentration and existing state of B in Si-Al melt (Fig. 15 a).
Solvent refining is an efficient method to remove B from MG-Si. Boride is usually formed and precipitated from the solvent by adding a small amount of additive to the solvent or alloy. The mechanism of B removal by solvent refining is based on the segregation behavior of borides between liquid and solid phases. The boride will be enriched in the liquid phase instead of staying in the purified silicon phase. The additives used included Ti, Zr, and Hf[93⇓⇓⇓~97][98,99][100]. The precipitation reaction of B in the solvent is represented by formula (25).
M e ( l ) + 2 B ( l ) = M e B 2 ( s )
Where Me refers to metallic Ti, Zr or Hf. The Ellingham plots of these borides are shown in Fig. 16. It is found that Ti, Zr and Hf show strong binding force with B in molten Si.
The results show that the solvent refining of Si-Al with Ti addition can effectively remove B from Si-Al melt. The solubility of TiB2 in Si-Al solvent is small, and B can be effectively removed by adding Ti to Si-Al solvent. The B content in the refined silicon is reduced from several hundred ppmw to several ppmw, while the residual amount of Ti is successfully controlled to about 1 ppmw[97]. A small amount of Zr or Hf is used as an additive to remove B by electromagnetic solidification of a Si-Al solvent[99][100]. When 1057 ppmw Zr was added, the removal rate of B in MG-Si was as high as 97.2%, and after treatment with HNO3-HF-H3PO4-HClO4 mixed acid, the removal rate of Zr added was as high as 99.7%[99]. Meanwhile, the addition of 1025 ppmw Hf also significantly increased the removal rate of B. When the liquidus temperature of Al-Si alloy is 1173 K and 410 ppmw of Hf is added, the removal rate of B in MG-Si is 98.2%, and the content of B in Si is 2.7 ppmw[100]. In addition, the results show that Zr and Hf are more effective in removing B than Ti, and the added Zr and Hf are also easier to remove than the added Ti[98].
As mentioned earlier, in the solvent refining process for the preparation of silicon for solar cell production, Ti, Zr and Hf can effectively remove B by forming borides, which solidify and bind in the eutectic phase and can be leached into the acidic solution. The selected additive should meet two requirements: (1) strong affinity for B; (2) The segregation coefficient between Si and Si-based melts is low. The former helps to enhance B removal, while the latter helps to remove additives through segregation behavior without causing serious pollution.
In addition, Chen et Al. Studied the mechanism of silicon extraction by adding a small amount of V to the Al-Si solvent to enhance the removal of B[101]. Si purification experiments were carried out by adding different amounts of V in Al-35 at% Si melt using an induction furnace. The results of Si purification show that a small amount of V can effectively improve the removal rate of B, and VSi2 is found in the eutectic Al-Si phase along the Si grain boundary. When the addition of V increased from 0 ppma to 1500 ppma, the content of B decreased from 73.6 ppma to 35.5 ppma, and the removal rate of B increased from 51.9% to 76.8%. The B content in refined Si can be significantly reduced by adding a proper amount of V during the Al-Si solvent refining process, because VB2 is formed during the Al-Si solvent refining process. Because of the small segregation coefficient of V and the small solubility of V in solid Si, the residual V content of refined Si was controlled at an undetectable level by ICP-OES, which means that the contamination of Si by V addition was eliminated during the refining process, thus clarifying the mechanism of purifying Si with V-added Al-Si solvent to remove B.

3.6.2 Silicon tin alloy

The Si-Sn system has attracted interest in the field of solvent refining because of the following advantages: (1) Sn is an electrically passive element in Si solar cells, and therefore, a trace amount of Sn (10 ppmw level) can be tolerated in refined Si; (2) Si-Sn alloys have a simple binary eutectic phase diagram, and Sn has limited solubility in Si and does not form intermetallic compounds[102][103]; (3) Due to the high vapor pressure of Sn, the Sn added to Si can be easily removed by vacuum treatment. In addition, Si-Sn is one of the most popular solvents for Si LPE.
In general, a single solvent refining process cannot achieve the desired results, and therefore, a combined purification system must be used to improve the purification efficiency. In this study, Si-Al and Si-Sn solvent refining were combined to refine MG-Si by considering their respective characteristics[104]. On the one hand, Si-Al alloys show higher Si solubility in Al and lower alloy melting point; On the other hand, the Si-Sn system is relatively efficient in separation because the density difference between Si(2.13 g/cm3) and Sn(7.01 g/cm3) is larger than that between Si and Al. The results show that P precipitates on the surface of primary Si in the form of P-containing phases such as Al-Ca-Mg-Si-P and Al-Si-P during solidification, and its final content can be reduced to 0.46 ppmw, with a removal rate of 99.8%. Similarly, binary silicides such as Fe3Si7 and Mn11Si19 or multi-component phases such as Ca-Mg-Si will be formed, which segregate on the grain boundary or adhere to the surface of Si, and can effectively remove metal impurities. In addition, their final removal rates were all above 99.9%.
In addition, a new process involving slag refining of Si-Sn alloy was studied as a way to enhance the boron removal effect of Si-Sn alloy[85]. It is mainly divided into the following four steps: (1) alloying of MG-Si with Sn to obtain Si-Sn alloy, (2) slag treatment, (3) solidification of Si-Sn melt, and (4) acid leaching to obtain primary Si. The distribution ratio of B (LB) increases significantly with the increase of Sn content in the alloy, which is attributed to the increase of activity coefficient of B and oxygen partial pressure, as shown in equations (26) ~ (29).
The overall reaction of B removal from molten Si by slag treatment is shown in equations (26) and (27).
B ( l ) + 3 / 4 S i O 2 ( l ) = B O 1.5 ( l ) + 3 / 4 S i ( l )
Δ G ° = 45490 - 24.1 T J / m o l
The molar distribution ratio of B is shown in equations (28) and (29).
K = ( α B O 1.5 ) ( α S i ) 3 / 4 α B ( α S i O 2 ) 3 / 4 = ( γ B O 1.5 X B O 1.5 ) γ B X B ( α S i α S i O 2 ) 3 / 4
L B = X B O 1.5 X B = K · ( γ B ) ( γ B O 1.5 ) ( α S i α S i O 2 ) 3 / 4
Under the condition that the slag composition is 40.5%CaO-35.5%SiO2-CaF2, the LB value increases from 2 to 200 when the Sn content increases from 5 to 82 mol%. In addition, the content of B finally decreased to 0.3 ppmw, as shown in Table 4[85]. These results indicate that B can be removed more effectively with a smaller amount of slag. Similar results were obtained for the purification of Si-Sn alloy by CaO-SiO2-Na2SiO3 system.
In order to effectively control the growth and enrichment of Si crystal, Ren et al. Studied a refining process using Si-Sn solvent system for directional solidification. In this process, Si-50 at% Sn melt was directionally solidified upward and downward at different moving rates, and the optimal conditions for crystal growth and enrichment were determined[105]. At the same time, a model was established to explain the growth process and enrichment mechanism of the crystal. The following conclusions are drawn. (1) The upward solidification process is more beneficial to the growth and enrichment of Si crystals than the downward solidification process: in the upward process, the amount of Si precipitation increases and it is easy to float up due to the buoyancy caused by the significant difference in density between the solidified Si and Si-Sn melts. (2) The enrichment ratio of Si increases with the decrease of the moving rate. The maximum enrichment ratio of Si reaches 98.6% when the upshift rate is 0.01 mm/min. (3) After HCl leaching, the residual Sn in the Si-rich zone can be easily reduced to the range of 0.00132 – 0.00163 at%, which is almost the same as the solid solubility range of Sn in Si.
In addition, Ren et al. Made a new attempt to remove B in Si by adding Zr as a scavenger in Si-Sn solvent solidification refining[106]. When Si-50 at% Sn melt (containing 120 ppmw boron) is used as refining solvent, the B removal rate increases from 36.9% to 73.6% with the increase of Zr content from 0 to 1055 ppma (6857 ppmw), which indicates that the addition of Zr is the main reason for the decrease of B content in refined Si. The removal rate of added Zr decreased slightly from 100% to 99.6% (26.6 ppmw in refined Si) with the increase of its initial content, because it is difficult for a larger amount of Zr to diffuse sufficiently from the Si grains into the intergranular and liquid phases in a limited time. The experimental results show that the B removal rate is improved to some extent by adding a small amount of Zr, and the refining process will be more effective under the conditions of lower cooling rate, upward moving direction and higher Zr content, as well as lower liquidus temperature. The maximum removal rate of B is 73.6%, and the added Zr is almost completely eliminated, which will not pollute the refined Si. However, due to the high liquidus temperature of Si-Sn alloy itself, the solubility product constant (Ksp) required for the formation of boride is large, which is not conducive to the formation and precipitation of boride, thus limiting the industrial application of this binary alloy system.

3.6.3 Si-Cu alloy

The process of purifying MG-Si with metal Cu has also been reported. The advantage of using this metal to purify MG-Si is that Si-Cu has a relatively low eutectic temperature, and Cu or its oxides in Si are easily removed by acid treatment[107].
Huang et al. Proposed the purification of silicon by the combination of Si-Cu solvent refining and acid leaching[108]. They first studied acid leaching with HCl, HNO3, aqua regia, and obtained the best effect of removing B and P by soaking Si-Cu powder in aqua regia, as shown in equations (30) and (31). It is worth noting that B and P precipitate in the form of MxBy and MxPy(M= metallic impurities), respectively, in the Si-Cu alloy.
M x B y + ( x y + 3 y ) / 3 H N O 3 ( a q ) + x y H C l ( a q ) + ( y - 2 / 3 x y ) H 2 O ( l ) → x M C l y ( a q ) + y B ( O H ) 3 ( a q ) + ( x y + 3 y ) / 3 N O ( g )
M x P y ( s ) + ( x y + 5 y ) / 3 H N O 3 ( a q ) + x y H C l ( a q ) + ( 2 y - 2 x y ) / H 2 O ( l ) → x M C l y ( a q ) + y H 3 P O 4 ( a q ) + ( x y + 5 y ) / 3 N O ( g )
In addition, the particle size, leaching agent, time, temperature and other parameters of Si-Cu alloy powder were comprehensively studied. The extraction yield of B and P increases with the increase of leaching temperature and leaching time, and the optimum particle size is 74 ~ 106 μm.
The combination of Si-Cu solvent refining and slag treatment can improve the removal rate of B and P in MG-Si. Li et al. Refined Si-Cu alloy with CaO-SiO2-Na2O-Al2O3 slag at 1773 K[62]. The refining results are shown in Fig. 17, which are consistent with those reported in the literature[36,109⇓⇓~112]. In the figure, the LB value increases linearly with the CaO/SiO2 ratio. Meanwhile, when the SiO2/Al2O3 ratio is 1.7, LB first increases to the maximum value (LB=46.8), and then decreases with further increase of the SiO2/Al2O3 ratio, indicating that B is easy to exist in the slag phase. It is worth noting that in the slag refining method of Si metal alloy, the use of Si-Cu alloy for slag refining is one of the effective methods to remove B.
Similarly, Huang et al. Used CaO-SiO2-CaCl2 slag to purify MG-Si in Si-Cu alloy[113]. The experimental results showed that the total removal rates of B and P in MG-Si increased from 71.5% to 88.7% and 32.7% to 57.6%, respectively, with the addition of Cu. Therefore, the increase in removal efficiency is mainly due to the higher impurity content in the Si-Cu alloy.
Ren et al. Proposed a new method to remove B from Si by Si-Cu solvent using Zr as a trapping agent[114]. At 1345 and 1258 K, the initial content of B was 460 and 310 ppma, respectively, and the obvious decrease of B and Zr content after equilibrium confirmed the removal effect of Zr addition on B. Electron probe microanalysis confirmed that the B-containing polygonal ZrBx precipitates found at the bottom of the specimen were ZrB2. Thermodynamic analysis shows that the solubility products of ZrB2 in Si-Cu melt are 2.5×10-12(1258 K,Si-57.0 at%Cu) and 4.11×10-12(1345 K,Si-50.0 at%Cu), respectively. Finally, by calculating the solubility product of ZrB2 and the equilibrium content between B and Zr at different liquidus temperatures, the composition of Si-Cu alloy that can form stable ZrB2 particles was estimated. This study provides direct evidence for the formation of ZrB2 by adding Zr to Si-Cu alloy to remove B, and also provides a theoretical basis for the subsequent directional solidification method to enrich large area of bulk polycrystalline Si with low B concentration.
According to the conclusion that B in Si-Cu melt can form ZrB2 precipitates by adding Zr, Ren et al. Carried out directional solidification refinement experiments in Si-Cu solvent with Zr addition, that is, using Si-Cu solvent to obtain large area of bulk silicon with low boron content[42]. The experimental results show that (1) the effect of boron removal is obviously improved, and the maximum is 93.4% (B: 80 → 5.3 ppmw), which is better than that of boron removal by adding Ti; (2) a large area of bulk Si is obtained by moving upward, and the maximum Si enrichment rate is 99.2%, which is higher than that of Si-Sn system; (3) The residual Cu content is reduced to a low level, with mole fractions between 0.00099 and 0.00191 (slightly higher than the solid solubility of Cu in Si). It can be seen that the segregation coefficient of B in Si-Cu system is small, and the solubility product of B and Zr is small, so a higher removal rate of B can be obtained.

3.6.4 Ferrosilicon

The natural abundance and lower production cost of Fe favor its use in solvent refining of Si. In addition, the solubility and segregation coefficient of Fe in Si are very low, which indicates that it is less likely to contaminate Si as a solvent.
Esfahani and Barati proposed a Si-Fe alloy system for refining MG-Si and systematically investigated the effect of cooling conditions (i.e., cooling rate and subsequent quenching temperature) on impurity segregation[39]. When the sample is quenched above the eutectic temperature (i.e., within the solid Si/liquid Fe-Si region), the removal of most impurities is improved compared to when quenched from the all-solid state (i.e., below the eutectic temperature). This result is mainly attributed to the diffusion of impurities to Si during the slow cooling in the second case. Wu et al. Explained that when Si-Fe alloy is quenched above the eutectic temperature, the temperature change will produce a sharp temperature gradient, which will enhance the segregation behavior of impurities in Si-Fe[115]. In addition, during the quenching process, more impurities were exposed and removed by acid leaching due to the crushing of the Si sheet. The elemental contents in Si after Si-Fe solvent refining are as follows: Al: 980 → 10 ppmw, B: 27 → 2 ppmw, Mn: 158 → 3 ppmw, Ni: 118 → 3 ppmw, Cr: 5 → 1 ppmw, Fe: 3108 → 1 ppmw, P: 68 → 29 ppmw. The concentrations of other impurities such as V, Ba, Li, Be, and Mg were below 0.5 ppmw. A large amount of Fe is alloyed with Si, but only 1 ppmw of Fe is detected in refined Si due to the extremely low segregation coefficient of Fe.
On the basis of the above experimental work, the thermodynamic study of B impurity in Si-Fe alloy was carried out, and the feasibility and effectiveness of B removal by solvent refining were confirmed[35,116]. The interaction parameters between B and Fe in solid Si are -813 ± 53 (1583 K), -830 ± 92 (1533 K) and -863 ± 91 (1483 K), respectively. The kB values at infinite dilution are 0.49 ± 0.01 (1583 K), 0.41 ± 0.03 (1533 K), and 0.33 ± 0.04 (1483 K). These values indicate a strong affinity between the two elements. In addition, the lower the temperature, the lower the distribution coefficient of B and the higher the yield of Si.

3.6.5 Other binary alloy systems.

Si-Na alloy: Morito et al. Investigated the Si-Na phase using differential thermal analysis and X-ray diffraction measurements[117]. As shown in Fig. 18, Si, as a primary crystalline phase, is clearly present in Si-rich composition at temperatures above 1023 K, thus indicating the possibility of formation and purification of Si from Si-Na alloy by evaporation of Na[118]. The concentrations of most impurity elements in the refined Si grains were analyzed by glow discharge mass spectrometry as follows: Fe: 3200 → 1.5 ppmw, Al: 510 → 6.0 ppmw, Ca: 96 → 1.7 ppmw, Ti: 170 → 0.016 ppmw, Cr: 520 → 1.6 ppmw, Mn: 260 → 0.49 ppmw, Ni: 61 → 0.18 ppmw. Certain transition metal silicides, such as FeSi2 and CrSi2, were observed between the bottom of the crucible and the lower surface of the Si grains. Na in refined Si is contaminated and its concentration increases from 2.6 ppmw to 85 ppmw, while Na in Si can be reduced to 0.055 ppmw under vacuum treatment due to the high Na vapor pressure at high temperature. In addition, the alloy also realizes the formation of single crystal, thin film and porous bulk Si under low temperature conditions[119,120].
Si-Ni alloy: Yin et al. Studied the effectiveness of Ni as a solvent to refine Si from both separation and purification[41]. The results show that a lower cooling rate leads to a higher impurity removal rate, and a longer holding time at the end of cooling is beneficial to the formation of lamellar crystals. The experimental results show that Al, Ba, Ca, Co, Cr, Fe, K, Mg, Mn, Mo, P, V, Zn and other elements in Mg-Si alloy are effectively removed from Si-Ni alloy. Therefore, Ni is considered to be a potential solvent-based metal.
Si-Ga alloys: Ga is a p-type dopant that can be alloyed with Si to form Si-Ga alloys[121]. According to the Si-Ga phase diagram, the alloy has a low eutectic temperature with only 0.006% eutectic Si[122]. As reported in the study, a Si epitaxial layer with a thickness of 30 ~ 50 μm was obtained from Si-Ga solution at 873 K[123]. Therefore, Ga can be used as a solvent for the purification of MG-Si.
Si-Ca alloy: Shimpo et al. Studied the interaction between Ca and P in molten Si and found that the alloying of Ca enhanced the segregation of P[83]. Therefore, the added Ca is an effective P absorber in molten MG-Si. The results show that the combination of Ca alloying refining and aqua regia leaching treatment has a high removal efficiency of P and Ca. He et al. Studied the removal of metal impurities in MG-Si by the combined treatment of these processes[124]. During the refining process of Ca alloying, the main impurity Ca precipitates in Si-Ca alloy or Si-Ca based alloy. By comparing the leaching effects of HCl, HF and aqua regia, it was found that HCl had the best removal effect on metal impurities in A-MG-Si. After continuous treatment, the contents of the main metal impurities Fe, Al, Ti and Ni were reduced to less than 5 ppmw. Based on these results, Meteleva-Fischer et al. Systematically summarized the effects of different parameters on the microstructure and impurity removal performance of Ca-alloyed MG-Si, and found that different Ca addition and cooling rate can affect the microstructure and impurity removal performance of Si[125]. In addition, Lai et al. Compared the leaching efficiency of Ca-alloyed MG-Si by analyzing the evolution of precipitates before and after the addition of HCl, aqua regia, and the mixture of HCl + HF, and determined that the mixture of HF + HCl was the best leaching agent for purifying MG-Si and A-MG-Si[126][127].
Although the above binary alloy systems can purify Si by removing impurities, the removal behavior of B element was not studied during the study.

3.6.6 Si-Al-Sn ternary alloy

As mentioned earlier, Si-Al alloys have advantages in B removal and lower refining temperatures (i.e., 1073 ~ 1473 K or 800 ~ 1200 ℃). However, it also has a disadvantage: specifically, most of the Si crystals remain in the eutectic phase (Si eutectic composition = 12.6 wt% at 850 K), so it is difficult to collect after solvent refining. In contrast, the eutectic Si content of the Si-Sn alloy is low (0.4 ppmw at 505 K), while the required Sn content in Si is high (10 ppmw), which can improve the recovery of primary Si. In order to overcome these shortcomings caused by the physical and chemical properties of the alloy system, Si-Al-Sn alloy, which combines Si-Al and Si-Sn alloy proposed by Tan's group, was used to purify MG-Si[121,128⇓⇓~131].
Xu et al. Calculated the interaction parameters of Sn and B at 1173 K to demonstrate the effect of Sn on B removal[130]. They found that there is a strong repulsive force between Sn and B with a value of ε B S n , S = 2506 ± 143 (1173 K). This finding is similar to that reported by Ma et al[132]. Therefore, the purification efficiency of Si-Al-Sn alloy will be weakened with the increase of Sn content. The experimental results of Fig. 19 confirm this point of view, in the Sn-free Si-Al alloy, the refining rate of B ( C S , B/ C O , B, where C S , B and C O , B are the B content of solid and initial mass, respectively) reaches the minimum value, but increases significantly with the increase of Sn content, so that the B removal ability of these alloys is obtained as: Si-Al alloy > Si-Al-Sn alloy > Si-Sn alloy[133]. In addition, the effect of Sn addition on the distribution of B in Si-Al-Sn alloy after MG-Si purification was also studied by electron probe microanalysis, and the results show that the intensity of B Kα line in the spectrum of primary Si phase is low, while the intensity of B Kα line in Al phase increases until reaching the highest level in Sn phase[87]. Therefore, the B content in Sn is much higher than that in Al and Si. This is mainly attributed to the segregation effect of B and the solidification sequence. During the solidification of Si-Al-Sn alloy, B is repelled by the growing solid phase, and its content in the remaining melt increases gradually. Because of the solidification sequence of Si-Al-Sn alloy (primary Si → eutectic Al → eutectic Sn), the content of B in the finally solidified Sn phase is higher.

3.6.7 Si-Al-Zn ternary alloy

Due to the low solubility of Si in Zn, which is similar to that in Sn, the Si-Al alloy with Zn addition was used as the solvent to refine MG-Si[134,135]. By studying the macrostructure of Si-Al alloys with different Zn addition, it is found that the shape of primary Si changes from elongated plate to irregular shape with the increase of Zn addition. Compared with the Si-63.8 wt% Al alloy (recovery 72.8 wt%), the recovery of the Si-Al-Zn alloy is improved by 21.3 wt% (recovery 94.1 wt%), thus greatly reducing the Si scrap.
Table 5 summarizes the reported efficiencies of B and P removal from different alloy systems. Compared with Si-Sn and other alloy systems, the boron removal efficiency of Si-Al alloy is higher under similar experimental conditions, which is suitable for large-scale industrial continuous processing.
Solvent refining has the advantages of high efficiency, environmental protection and easy to carry out at lower temperature, but its biggest problem is the effective separation of primary Si and eutectic Si. Of course, there seem to be many ways to solve this problem, such as hypergravity, the addition of electromagnetic fields, and directional solidification. However, none of these can be called an efficient pathway for Si segregation. In addition, the introduction of metallic impurities may overload the directional solidification process, bringing more refinement tasks. And the alloying of Si will lead to an increase in production costs, and the separation of alloys is also a challenge, for example, the density of Si-Al alloy and Al is not much different, how to effectively separate Si-Al alloy after purification. At present, the recycling of solvents is also a thorny problem for researchers, and the removal of impurities in alloys requires a large amount of acid, which will cause the loss of some raw materials.
Solvent refining can be used for large-scale production of pure Si, and low temperature operation is beneficial to reduce the cost. However, further research is needed to determine the best solvent for optimal purification efficiency, minimizing solvent contamination, and adequate separation of refined Si from the solvent.

3.7 Application of boron removal proces by metallurgical method

No matter which of the above technologies is used alone, it can not remove all impurities at the same time, often only one or several of them are effective, and even the removal of a specific impurity can not meet the requirements of use. Therefore, in the purification process of Si, the above technologies are not used independently, but in combination, so the metallurgical process is a composite purification process that integrates multiple methods.

3.7.1 PM process

Shanghai Puluo Company adopts the self-developed PM purification technology to prepare solar-grade polycrystalline Si. The process can start from silica ore, through improved metal Si smelting method, secondary refining, directional solidification, DVC double vacuum purification technology and other technologies, with metal Si with purity of 2 ~ 3 N as raw material, solar-grade polycrystalline Si with purity of 6.3 ~ 6.7 N can be obtained. The process route is roughly as follows: local vacuum of concentrate → slagging → powder metallurgy → hydrometallurgy → gas blowing refining → vacuum melting and directional solidification → ingot testing. The process produces solar-grade polycrystalline Si with very low energy consumption and very little pollution, and the unit energy consumption can be reduced to less than 12 kW · H/kg. The produced polycrystalline Si product has high purity, excellent resistivity and long minority carrier lifetime, so the photoelectric conversion efficiency is high and the attenuation is small.

3.7.2 Elkem process

Elkem is a Norwegian company that produces metallurgical grade silicon and is the largest producer of metallurgical grade silicon in the world. The process route for producing the solar grade silicon by the metallurgical method comprises the following steps of: selecting pure metallurgical grade silicon; carrying out slagging treatment; carrying out wet treatment; carrying out directional solidification; slicing; polishing; and chemically leaching[150]. The slag former and Si liquid are mixed to remove B, then the metal impurities in the granular Si are removed by pickling, and the impurities are removed by directional solidification in a special device. The concentration of B in the raw material metallurgical grade silicon is up to 300×10-6,P and up to 100×10-6. After the above refining treatment, the concentration of B in the final solar grade silicon is 0.2×10-6~10×10-6,P and 0.1×10-6~10×10-6, and the concentration of metal impurities is below 150×10-6. The process flow is simple and the production cost is low, but the light attenuation of the product is too large. At present, Elkem is also actively solving this problem.

3.7.3 Kawasaki Steel process

With the support of NEDO, Kawasaki Steel Company in Japan has developed a process route for the preparation of solar-grade silicon by metallurgical method, which is divided into two stages. The first stage is carried out in an absolute vacuum environment, metallurgical grade silicon is used as raw material, and the raw material is injected into a crucible in a melting furnace with an electron beam generating device;Vacuum distillation technology is adopted to volatilize P and other volatile impurities from the surface of the Si melt so as to discharge the impurities out of the furnace, and after the process is continued for a period of time, the Si melt is directionally solidified for the first time to remove metal impurities; In the second stage, the Si material obtained in the first stage is used as a raw material and is melted in a plasma melting furnace, and mixed gas of water vapor, hydrogen and argon is introduced at the same time, so that impurities such as B, C and the like on the surface of the Si melt are removed, and the treated Si melt is directionally solidified for the second time to further remove metal impurities[151]. The experimental results show that vacuum evaporation and plasma oxidation can reduce the concentration of B and P impurities to the level of 0.1×10-6, and two directional solidification can also make the metal impurities meet the requirements of solar grade silicon, and the conversion efficiency of solar cells made of polycrystalline silicon produced by this process can also reach 14. 1%. However, the electron beam, plasma gun and other devices make the equipment more complex, and the twice melting and directional solidification treatment greatly increases the production cost and reduces the production efficiency. Aiming at the problems existing in the development of the project, Professor Kazuki Morita of Tokyo University in Japan found that the addition of Ca is beneficial to the removal of Fe and Ti impurities in metal Si through thermodynamic research.

3.7.4 Vacuum metallurgical process

Professor Ma Wenhui of the National Engineering Laboratory of Vacuum Metallurgy, Kunming University of Science and Technology, proposed a new process for preparing solar grade silicon by vacuum metallurgy based on the impurity distribution characteristics of metallurgical grade silicon[152]. The process combines new technologies such as vacuum drying, vacuum refining, vacuum distillation, vacuum degassing and vacuum directional solidification to directly prepare solar grade silicon from metallurgical grade silicon, and the purity of silicon products exceeds 99.99%. At the same time, the National Engineering Laboratory of Vacuum Metallurgy of Kunming University of Science and Technology and the International Research Center for Sustainable Materials of the University of Tokyo have cooperated to carry out in-depth and systematic research, continuously improve and optimize the existing technology, improve the removal efficiency of impurities in Si, meet the requirements of solar grade silicon, and promote the realization of large-scale industrialization.

4 Conclusion and prospect

Slag treatment is still the main refining method for removing B from Si. In slag treatment, oxidation reaction is the main mechanism reaction for B removal. It is well known that the main component of slag is silicate, while the third component or multi-component (such as MgO, Li2O, Al2O3, CaF2) can only change the properties of slag. To optimize the slag treatment process, the introduction of Na2O/K2O/CaCl2 and its combination with other methods was found to be very effective. In the production of MG-Si, oxygen or gas blowing has been successfully used to remove metal impurities such as Fe, Al and Ca, but the removal effect of gas blowing on non-metal impurities such as B and P is not obvious. However, in the gas blowing refining process, blowing wet oxygen (mixed gas of water vapor and O2) or wet hydrogen (mixed gas of water vapor and H2) into the Si melt for refining can indeed significantly improve the B removal efficiency. In addition, the combined refining process of gas blowing and slag treatment can significantly improve the B removal efficiency. The secondary refining technology of SOG-Si production by metallurgical method is the key step of B removal, and the product after B removal by secondary refining can meet the requirements of solar cells for B content.
Plasma treatment has the advantages of high refining temperature, high chemical reactivity and strong degassing ability. Plasma refining has a good removal effect on Al, Ca, Fe and B, but it also has the disadvantages of expensive production equipment, high energy consumption and intermittent production, which limits the development of its industrial application.
Acid leaching treatment can effectively remove Fe, Al, Ca, Ti and other metal impurities in Si, and also promote the removal of B and P impurities. And the solvent metal and acid can be converted into valuable by-products, so the cost of using the acid leaching process in industrial production is very low.
Solvent refining is an emerging technology to remove impurities to produce SOG-Si, which has attracted more and more attention. It is a purification process of Si recrystallization from supersaturated melt according to the segregation behavior of different elements. MG-Si is used as raw material, and other elements are added in the solvent refining process to prepare the alloy, so that the segregation coefficient of the target impurity is effectively reduced, and the impurity is easier to stay in the molten alloy phase along with the solidification of Si, so that the removal efficiency of the impurity in Si is improved.
Solvent refining can effectively remove most of the metal impurities in Si, and has certain advantages in removing impurities B and P. The selection of solvent metal for solvent refining purification of MG-Si mainly depends on the properties of impurities in Si. The appropriate metal solvent should be selected according to the specific properties of different impurities in Si melt to realize the selective removal of impurity elements, especially the enhanced precipitation of non-metallic impurities B and P, so as to improve the purification efficiency of MG-Si and realize the quantitative production of solar grade silicon at low cost.
Combined with the principle of solvent refining, it can be found that it has significant advantages: (1) low energy consumption: it can be applied at a lower temperature (400 K below the melting point of Si), which can effectively reduce energy consumption; (2) The removal rate of B is high: the segregation coefficient of B will be reduced, which is more conducive to the removal of B; (3) The thermodynamic properties of B are reshaped: the recrystallization of primary Si in the melt of Si-based alloy can lead to the redistribution of impurity B, which is convenient to reshape the occurrence and distribution behavior of B, and can also control the enrichment of primary Si; (4) It is easy to strengthen the refining process: compared with the pure Si system, the properties of B in the Si-based alloy melt have changed, and the B-philic element additive can be used to strengthen the refining process. If the boron removal process can be improved by using a small amount of additives with strong affinity to B to enhance the formation and precipitation of boride, it will be more effective to remove B, increase the practicability of the process, reduce the impurity content, and finally meet the SOG-Si standard. Based on the advantages of solvent refining method, this method is one of the most potential methods to achieve low-cost quantitative production of solar grade silicon, and has a good development prospect.
With the rapid increase of energy demand and the continuous reduction of photovoltaic power generation cost, the development of photovoltaic industry and the demand for Si materials will show a steady upward trend in the future. With the advantages of high efficiency, low cost, low energy consumption and environmental friendliness, metallurgical method will still be the research focus and focus in this field, which will promote the innovation in the field of polycrystalline Si purification and continuously promote the rapid development of photovoltaic industry.

References

[1]

I.S.E. Fraunhofer, Photovoltaics report. Fraunhofer ISE, Freiburg, 2020: 4.

[2]

Shimoda T, Masuda T. Jpn. J. Appl. Phys., 2014, 53(2S): 02BA01.

[3]

Tang K, Øvrelid E J, Tranell G, Tangstad M. Mater. Trans., 2009, 50(8): 1978.

[4]

Samsonov G V, Sleptsov V M. Russ. J. Inorg. Chem., 1963, 8: 1047.

[5]

Brosset C, Magnusson B. Nature, 1960, 187(4731): 54.

[6]

Trumbore F A, Freeland P E, Logan R A. J. Electrochem. Soc., 1961, 108(5): 458.

[7]

Hesse J. Z. Met., 1968, 59: 499.

[8]

Olesinski R W, Abbaschian G J. Bull. Alloy Phase Diagr., 1984, 5(5): 476.

[9]

Fuller C S, Ditzenberger J A. J. Appl. Phys., 1956, 27(5): 544.

[10]

Garandet J P. Int. J. Thermophys., 2007, 28(4): 1285.

[11]

Kodera H. Jpn. J. Appl. Phys., 1963, 2(4): 212.

[12]

Williams E L. J. Electrochem. Soc., 1961, 108(8): 795.

[13]

Barry M L, Olofsen P. J. Electrochem. Soc., 1969, 116(6): 854.

[14]

Fuller C S, Ditzenberger J A. J. Appl. Phys., 1954, 25(11): 1439.

[15]

Okamura M. Jpn. J. Appl. Phys., 1969, 8(12): 1440.

[16]

Domínguez E, Jaraiz M. J. Electrochem. Soc., 1986, 133(9): 1895.

[17]

Vick G L, Whittle K M. J. Electrochem. Soc., 1969, 116(8): 1142.

[18]

Wijaranakula W. Jpn. J. Appl. Phys., 1993, 32(Part 1, No. 9A): 3872.

[19]

Ghoshtagore R N. Phys. Rev. B, 1971, 3(2): 397.

[20]

Tang K, Øvrelid E J, Tranell G, Tangstad M. Thermochemical and Kinetic Databases for the Solar Cell Silicon Materials, Crystal Growth of Si for Solar Cells. Springer, Berlin, Heidelberg, 2009. 219.

[21]

Noguchi R, Suzuki K, Tsukihashi F, Sano N. Metall. Mater. Trans. B, 1994, 25(6): 903.

[22]

Fang M, Lu C H, Huang L Q, Lai H X, Chen J, Yang X B, Li J T, Ma W H, Xing P F, Luo X T. Ind. Eng. Chem. Res., 2014, 53(30): 12054.

[23]

Krystad E, Tang K, Tranell G. JOM, 2012, 64(8): 968.

[24]

Zhang L, Tan Y, Li J Y, Liu Y, Wang D. Mater. Sci. Semicond. Process, 2013, 16(6): 1645.

[25]

Tang K, Øvrelid E J, Tranell G, Tangstad M. J. Occup. Med., 2009, 61(11): 49.

[26]

Wu J J, Wang F M, Chen Z J, Ma W H, Li Y L, Yang B, Dai Y N. Fluid Phase Equilibria, 2015, 404: 70.

[27]

Wu J J, Wang F M, Ma W H, Lei Y, Yang B. Metall. Mater. Trans. B, 2016, 47(3): 1796.

[28]

Nishimoto H, Kang Y, Yoshikawa T, Morita K. High Temp. Mater. Process., 2012, 31(4/5): 471.

[29]

Wang Y, Morita K. J. Sustain. Metall., 2015, 1(2): 126.

[30]

Wang F M, Wu J J, Ma W H, Lei Y, Wei K X, Yang B. J. Chem. Thermodyn., 2018, 118: 215.

[31]

Tanahashi M, Fujisawa T, Yamauchi C. Shigen-to-Sozai, 1998, 114(11): 807.

[32]

Inoue G, Yoshikawa T, Morita K. High Temp. Mater. Process., 2003, 22(3/4): 221.

[33]

Yoshikawa T, Morita K. Mater. Trans., 2005, 46(6): 1335.

[34]

Yoshikawa T, Morita K. Metall. Mater. Trans. B, 2005, 36(6): 731.

[35]

Khajavi L T, Morita K, Yoshikawa T, Barati M. J. Alloys Compd., 2015, 619: 634.

[36]

Teixeira L A V, Morita K. ISIJ Int., 2009, 49(6): 783.

[37]

Xu F M, Wu S R, Tan Y, Li J Y, Li Y Q, Liu Y. Sep. Sci. Technol., 2014, 49(2): 305.

[38]

Hall R N. J. Phys. Chem., 1953, 57(8): 836.

[39]

Esfahani S, Barati M. Met. Mater. Int., 2011, 17(6): 1009.

[40]

Zhao L X, Wang Z, Guo Z C, Li C Y. T. Nonfree. Metal. Soc., 2011, 21(5): 1185.

[41]

Yin Z, Oliazadeh A, Esfahani S, Johnston M, Barati M. Can. Metall. Q., 2011, 50(2): 166.

[42]

Ren Y S, Morita K. ACS Sustain. Chem. Eng., 2020, 8(17): 6853.

[43]

Khattak C P, Joyce D B. National Renewable Energy Laboratory, 2001. 8.

[44]

Sortland Ø S, Tangstad M. Metall. Mater. Trans. E, 2014, 1(3): 211.

[45]

Khattak C P, Joyce D B, Schmid F. Sol. Energy Mater. Sol. Cells, 2002, 74(1/4): 77.

[46]

Nordstrand E F, Tangstad M. Metall. Mater. Trans. B, 2012, 43(4): 814.

[47]

Altenberend J, Chichignoud G, Delannoy Y. Metall. Mater. Trans. E, 2017, 4(1): 41.

[48]

Lee B P, Lee H M, Park D H, Shin J S, Yu T U, Moon B M. Sol. Energy Mater. Sol. Cells, 2011, 95(1): 56.

[49]

Flamant G, Kurtcuoglu V, Murray J, Steinfeld A. Sol. Energy Mater. Sol. Cells, 2006, 90(14): 2099.

[50]

Wu J J, Ma W H, Yang B, Dai Y N, Morita K. Trans. Nonferrous Met. Soc. China, 2009, 19(2): 463.

[51]

Wu J J, Li Y L, Ma W H, Liu K, Wei K X, Xie K Q, Yang B, Dai Y N. Silicon, 2014, 6(1): 79.

[52]

Suzuki K, Kumagai T, Sano N. ISIJ Int., 1992, 32(5): 630.

[53]

Jones J A T, Bowman B, Lefrank P A. Making, Shaping, and Treating of Steel. Steelmaking and Refining Volume, The AISE Steel Foundation, Pittsburgh, PA, 1998. 281.

[54]

Jung E J, Moon B M, Seok S H, Min D J. Energy, 2014, 66: 35.

[55]

Wu J J, Li Y L, Ma W H, Wei K X, Yang B, Dai Y N. Trans. Nonferrous Met. Soc. China, 2014, 24(4): 1231.

[56]

Jakobsson L K, Tangstad M. Metall. Mater. Trans. B, 2015, 46(2): 595.

[57]

Luo D W, Liu N, Lu Y P, Zhang G L, Li T J. Trans. Nonferrous Met. Soc. China, 2011, 21(5): 1178.

[58]

Li M, Utigard T, Barati M. Metall. Mater. Trans. B, 2015, 46(1): 74.

[59]

Johnston M D, Barati M. J. Non Cryst. Solids, 2011, 357(3): 970.

[60]

Jakobsson L K, Tangstad M. Metall. Mater. Trans. B, 2014, 45(5): 1644.

[61]

Li Y L, Wu J J, Ma W H, Yang B. Silicon, 2015, 7(3): 247.

[62]

Li M, Utigard T, Barati M. Metall. Mater. Trans. B, 2014, 45(1): 221.

[63]

Zhang L, Tan Y, Xu F M, Li J Y, Wang H Y, Gu Z. Sep. Sci. Technol., 2013, 48(7): 1140.

[64]

Zhang L, Tan Y, Li J Y, Liu Y, Wang D K. Mater. Sci. Semicond. Process., 2013, 16(6): 1645.

[65]

Safarian J, Tranell G, Tangstad M. Metall. Mater. Trans. B, 2013, 44(3): 571.

[66]

Yin C H, Hu B F, Huang X M. J. Semicond., 2011, 32(9): 092003.

[67]

White J F, Du S C. Metall. Mater. Trans. B, 2014, 45(1): 96.

[68]

Wu J J, Xia Z F, Ma W H, Wang F M, Zhou Y Q, Wei K X. Mater. Sci. Semicond. Process., 2017, 57: 59.

[69]

Wang F M, Wu J J, Ma W H, Xu M, Lei Y, Yang B. Sep. Purif. Technol., 2016, 170: 248.

[70]

Wu J J, Zhou Y Q, Ma W H, Xu M, Yang B. Metall. Mater. Trans. B, 2017, 48(1): 22.

[71]

Xia Z F, Wu J J, Ma W H, Lei Y, Wei K X, Dai Y N. Sep. Purif. Technol., 2017, 187: 25.

[72]

Johnston M D, Khajavi L T, Li M, Sokhanvaran S, Barati M. JOM, 2012, 64(8): 935.

[73]

Benmansour M, Nikravech M, Morvan D, Amouroux J, Chapelle J. J. Phys. D: Appl. Phys., 2004, 37(21): 2966.

[74]

Rousseau S, Benmansour M, Morvan D, Amouroux J. Sol. Energy Mater. Sol. Cells, 2007, 91(20): 1906.

[75]

Nakamura N, Baba H, Sakaguchi Y, Kato Y. Mater. Trans., 2004, 45(3): 858.

[76]

Yuge N, Baba H, Sakaguchi Y, Nishikawa K, Terashima H, Aratani F. Sol. Energy Mater. Sol. Cells, 1994, 34(1/4): 243.

[77]

Yuge N, Abe M, Hanazawa K, Baba H, Nakamura N, Kato Y, Sakaguchi Y, Hiwasa S, Aratani F. Prog. Photovolt: Res. Appl., 2001, 9(3): 203.

[78]

Díez M D, Fjeld M, Andersen E, Lie B. Chem. Eng. Sci., 2006, 61(1): 229.

[79]

Pizzini S. Sol. Energy Mater., 1982, 6(3): 253.

[80]

Dietl J. Sol. Cells, 1983, 10(2): 145.

[81]

Sun Y H, Ye Q H, Guo C J, Chen H Y, Lang X, David F, Luo Q W, Yang C M. Hydrometallurgy, 2013, 139: 64.

[82]

Lai H X, Huang L Q, Gan C H, Xing P F, Li J T, Luo X T. Hydrometallurgy, 2016, 164: 103.

[83]

Shimpo T, Yoshikawa T, Morita K. Metall. Mater. Trans. B, 2004, 35(2): 277.

[84]

Yoshikawa T, Morita K. J. Electrochem. Soc., 2003, 150(8): G465.

[85]

Ma X D, Yoshikawa T, Morita K. Metall. Mater. Trans. B, 2013, 44(3): 528.

[86]

Mitrašinović A M, Utigard T A. Silicon, 2009, 1(4): 239.

[87]

Li Y Q, Tan Y, Li J Y, Xu Q, Liu Y. J. Alloys Compd., 2014, 583: 85.

[88]

Zhao L X, Wang Z, Guo Z C, Li C Y. Trans. Nonferrous Met. Soc. China, 2011, 21(5): 1185.

[89]

Yoshikawa T, Morita K. J. Cryst. Growth, 2009, 311(3): 776.

[90]

Yoshikawa T, Morita K. Sci. Technol. Adv. Mater., 2003, 4(6): 531.

[91]

Zou Q C, Jie J C, Sun J L, Wang T M, Cao Z Q, Li T J. Sep. Purif. Technol., 2015, 142: 101.

[92]

Li Y L, Chen J, Dai S Y. J. Cryst. Growth, 2016, 453: 49.

[93]

Ban B Y, Li J W, Bai X L, He Q X, Chen J, Dai S Y. J. Alloys Compd., 2016, 672: 489.

[94]

Bai X L, Ban B Y, Li J W, Fu Z Q, Peng Z J, Wang C B, Chen J. Sep. Purif. Technol., 2017, 174: 345.

[95]

Li Y L, Chen J, Ban B Y, Zhang T T, Dai S Y. High Temp. Mater. Process., 2015, 34(1): 43.

[96]

Ban B Y, Li Y L, Zuo Q X, Zhang T T, Chen J, Dai S Y. J. Mater. Process. Technol., 2015, 222: 142.

[97]

Yoshikawa T, Arimura K, Morita K. Metall. Mater. Trans. B, 2005, 36(6): 837.

[98]

Lei Y, Ma W H, Sun L E, Wu J J, Morita K. Sep. Purif. Technol., 2016, 162: 20.

[99]

Lei Y, Ma W H, Sun L E, Dai Y N, Morita K. Metall. Mater. Trans. B, 2016, 47(1): 27.

[100]

Lei Y, Ma W H, Sun L E, Wu J J, Dai Y N, Morita K. Sci. Technol. Adv. Mater., 2016, 17(1): 12.

[101]

Chen K, Chen X H, Lei Y, Ma W H, Han J X, Yang Z H. Sep. Purif. Technol., 2018, 203: 168.

[102]

Hopkins R H, Rohatgi A. J. Cryst. Growth, 1986, 75(1): 67.

[103]

Olesinski R W, Abbaschian G J. J. Phase Equilib., 1984, 5(3): 273.

[104]

Hu L, Wang Z, Gong X Z, Guo Z C, Zhang H. Metall. Mater. Trans. B, 2013, 44(4): 828.

[105]

Ren Y S, Wang H P, Morita K. Vacuum, 2018, 158: 86.

[106]

Ren Y S, Wang H P, Morita K. Vacuum, 2019, 167: 319.

[107]

Olesinski R W, Abbaschian G J. Bull. Alloy Phase Diagr., 1986, 7(2): 170.

[108]

Huang L Q, Lai H X, Lu C H, Fang M, Ma W H, Xing P F, Li J T, Luo X T. Hydrometallurgy, 2016, 161: 14.

[109]

Cai J, Li J T, Chen W H, Chen C, Luo X T. Trans. Nonferrous Met. Soc. China, 2011, 21(6): 1402.

[110]

Jakobsson L K, Tangstad M. Metall. Mater. Trans. B, 2014, 45(5): 1644.

[111]

Zhang L, Tan Y, Xu F M, Li J Y, Wang H Y, Gu Z. Sep. Sci. Technol., 2013, 48(7): 1140.

[112]

Suzuki K, Sano N, Tenth E C. Photovoltaic Solar Energy Conference. Dordrecht: Springer Netherlands, 1991. 273/275.

[113]

Huang L Q, Lai H X, Gan C H, Xiong H P, Xing P F, Luo X T. Sep. Purif. Technol., 2016, 170: 408.

[114]

Ren Y S, Ueda S, Morita K. ACS Sustain. Chem. Eng., 2019, 7(24): 20107.

[115]

Wu J J, Wang Z, Hu X J, Guo Z C. Chin. J. Process. Eng., 2014, 14: 64.

[116]

Khajavi L T, Morita K, Yoshikawa T, Barati M. Metall. Mater. Trans. B, 2015, 46(2): 615.

[117]

Morito H, Yamada T, Ikeda T, Yamane H. J. Alloys Compd., 2009, 480(2): 723.

[118]

Morito H, Karahashi T, Uchikoshi M, Isshiki M, Yamane H. Silicon, 2012, 4(2): 121.

[119]

Morito H, Karahashi T, Yamane H. J. Cryst. Growth, 2012, 355(1): 109.

[120]

Yamane H, Morito H, Uchikoshi M. J. Cryst. Growth, 2013, 377: 66.

[121]

Li J W, Ban B Y, Li Y L, Bai X L, Zhang T T, Chen J. Silicon, 2017, 9(1): 77.

[122]

Olesinski R W, Kanani N, Abbaschian G J. Bull. Alloy Phase Diagr., 1985, 6(4): 362.

[123]

Girault B, Chevrier F, Joullie A, Bougnot G. J. Cryst. Growth, 1977, 37(2): 169.

[124]

He F L, Zheng S S, Chen C. Metall. Mater. Trans. B, 2012, 43(5): 1011.

[125]

Meteleva-Fischer Y V, Yang Y X, Boom R, Kraaijveld B, Kuntzel H. Miner. Process. Extr. Metall., 2013, 122(4): 229.

[126]

Lai H X, Huang L Q, Lu C H, Fang M, Ma W H, Xing P F, Li J T, Luo X T. Hydrometallurgy, 2015, 156: 173.

[127]

Xi F S, Li S Y, Ma W H, Chen Z J, Wei K X, Wu J J. Hydrometallurgy, 2021, 201: 105553.

[128]

Li J Y, Cao P P, Ni P, Li Y Q, Tan Y. Sep. Sci. Technol., 2016, 51(9): 1598.

[129]

Li J Y, Liu Y, Tan Y, Li Y Q, Zhang L, Wu S R, Jia P J. J. Cryst. Growth, 2013, 371: 1.

[130]

Xu F M, Wu S R, Tan Y, Li J Y, Li Y Q, Liu Y. Sep. Sci. Technol., 2014, 49(2): 305.

[131]

Li Y Q, Li J Y, Tan Y, Zhang L F, Kazuki M. J. Inorg. Mater., 2016, 31(8): 791.

[132]

Ma X D, Yoshikawa T, Morita K. J. Alloys Compd., 2012, 529: 12.

[133]

Li Y Q. Doctoral Dissertation of Dalian University of Technology, 2015.

[134]

(李亚琼. 大连理工大学博士论文, 2015.).

[135]

Li Y Q, Tan Y, Li J Y, Morita K. J. Alloys Compd., 2014, 611: 267.

[136]

Li Y, Tan Y, Cao P, Li J, Jia P, Liu Y. Mater. Res. Innov., 2015, 19(2): 81.

[137]

Obinata I, Komatsu N. J. Jpn. Inst. Met. Mater., 1954, 18(5): 279.

[138]

Yoshikawa T, Morita K. ISIJ Int., 2005, 45(7): 967.

[139]

Gumaste J L, Mohanty B C, Galgali R K, Syamaprasad U, Nayak B B, Singh S K, Jena P K. Sol. Energy Mater., 1987, 16(4): 289.

[140]

Gu X, Yu X G, Yang D R. Sep. Purif. Technol., 2011, 77(1): 33.

[141]

Li J W, Guo Z C, Tang H Q, Wang Z, Sun S T. Trans. Nonferrous Met. Soc. China, 2012, 22(4): 958.

[142]

Jie J C, Zou Q C, Wang H W, Sun J L, Lu Y P, Wang T M, Li T J. J. Cryst. Growth, 2014, 399: 43.

[143]

Xue H Y, Lv G Q, Ma W H, Chen D T, Yu J. Metall. Mater. Trans. A, 2015, 46(7): 2922.

[144]

Fang M, Lu C H, Lai H X, Huang L Q, Chen J, Ma W H, Sheng Z L, Shen J N, Li J T, Luo X T. Mater. Sci. Technol., 2013, 29(7): 861.

[145]

Zhang L F, Ma Y S, Li Y Q. Mat. Sci. Semicon. Proc., 2017, 71: 12.

[146]

Ma X D, Lei Y, Yoshikawa T, Zhao B J, Morita K. J. Cryst. Growth, 2015, 430: 98.

[147]

Ma X D, Yoshikawa T, Morita K. Sep. Purif. Technol., 2014, 125: 264.

[148]

Hu L, Wang Z, Gong X Z, Guo Z C, Zhang H. Sep. Purif. Technol., 2013, 118: 699.

[149]

Li J Y, Jia P J, Li Y Q, Cao P P, Liu Y, Tan Y. J. Mater. Sci. Mater. Electron., 2014, 25(4): 1751.

[150]

Lei Y, Ma W H, Lv G Q, Wei K X, Li S Y, Morita K. Sep. Purif. Technol., 2017, 173: 364.

[151]

Zhang J, Li T J, Ma X D, Luo D W, Liu N, Liu D H. J. Semicond., 2009, 30(5): 053002.

[152]

Rousseau S, Benmansour M, Morvan D, Amouroux J. Sol. Energy Mater. Sol. Cells, 2007, 91(20): 1906.

[153]

Ma W H, Dai Y N, Yang B, Liu D C, Wang H. Advanced Materials Industry, 2006, (10): 12.

[154]

(马文会, 戴永年, 杨斌, 刘大春, 王华. 新材料产业, 2006, (10): 12.).

Funding

Youth Fund of the National Natural Science Foundation of China(52104303)

Yunnan Joint Fund of the National Natural Science Foundation of China(U1702251)

Open Fund of the State Key Laboratory of Advanced Metallurgy of University of Science and Technology Beijing(KK21-07)

PDF (13119KB)

1730

Accesses

0

Citation

Detail

Sections
Recommended

/

〈 〉